CN103761668A - Method and system for detecting bad users in network transaction - Google Patents

Abstract

The invention discloses a method and system for detecting bad users in a network transaction. The method for detecting the bad users in the network transaction comprises the steps that active period detection is carried out, wherein the active period of at least one transaction object is detected on the basis of historical information; transaction fraud detection is carried out, wherein the active periods of the transaction objects are detected on the basis of at least one piece of evidence, and transaction objects with transaction ranking fraud are determined; bad user detection is carried out, wherein the bad users which execute at least one transaction operation in the active periods of the transaction objects with the transaction ranking fraud are detected. According to the method and system for detecting the bad users in the network transaction, the bad users who possibly participate in fraud behaviors in the network transaction can be obtained through detection, the important foundation is provided for all the network transaction users to obtain real information of commodities or services, and safety of the network transaction is improved.

Description

Bad user's detection method and detection system in network trading

Technical field

The present invention relates to network field, relate in particular to detection method and the system of bad user in a kind of network trading.

Background technology

The transaction of carrying out commodity or service by networks such as internets is more and more subject to the network user's welcome, and this makes network trading day by day become a kind of important industry.In order to facilitate user to select and buy commodity or service by network, a lot of e-commerce websites can intensively provide the services such as the displaying, inquiry, evaluation of commodity or service, simultaneously also can be termly, for example every day, the ranking list (Leader Board) of publishing commodity or service is to embody some current commodity popular with users or services.In fact; this ranking list is one of most important means of commodity sales promotion or service; commodity or service very high rank in ranking list can stimulate a large amount of concerns of user even to buy these commodity or service conventionally, and bring huge economic return for the seller of commodity or service.Therefore, the seller of commodity or service wishes that its commodity or service occupy higher rank in ranking list very much.

The transaction rank swindle (Ranking Fraud) of commodity or service refers to that object is to improve rank in network ranking list of commodity or service and the transaction deceptive practices carried out.In fact, be different from the traditional market means of dependence and improve the rank of commodity or service, the seller of commodity or service by exaggerating its sales volume or issue false user to evaluate to implement the behavior of rank swindle more and more general, for example, employs " waterborne troops (human water armies) " so bad user promote at short notice the purchase volume of commodity or service and evaluate number of times etc.

In order to make the potential buyer of commodity or service can obtain the real information of commodity or service, be necessary to identify the bad user of these rank swindles of participating in business, special processing etc. is rejected or is done in the transaction operation just likely these bad users being participated on this basis, thereby improves the security of network trading.

In the prior art, for bad user's identification be transaction Network Based user's transactions history record and evaluate other people/consistance of thing etc. infers.But, this transactions history record and evaluate other people/the conforming judgement of thing often needs artificial participation, takes time and effort, and often owing to not checking global data, cannot make judgement accurately.

Summary of the invention

The object of the present invention is to provide the detection technique of bad user in a kind of network trading, for identifying the bad user of possibility participation network transaction fraud, the real information that obtains commodity or service for all-network transaction user provides important foundation, improves the security of network trading.

For solving the problems of the technologies described above, according to an aspect of the present invention, provide the detection method of bad user in a kind of network trading, described method comprises:

Active period detecting step, detects the active period of at least one trading object based on historical information;

Transaction swindling detecting step, detects the active period of described at least one trading object based at least one evidence, determines the trading object that has the swindle of transaction rank;

Bad user's detecting step, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.

According to another aspect of the present invention, also provide the detection system of bad user in a kind of network trading, described system comprises:

Active period detecting unit, for detecting the active period of at least one trading object based on historical information;

Transaction swindling detecting unit, for based at least one evidence, the active period of described at least one trading object being detected, determines the trading object that has the swindle of transaction rank;

Bad user's detecting unit, carries out at least one in the active period for detection of the trading object of swindling in described existence transaction rank and hands over easy-operating bad user.

According to another aspect of the present invention, also provide the detection method of bad user in a kind of network trading, described method comprises:

Transaction swindling detecting step, detects the active period of at least one trading object based at least one evidence, determines the trading object that has the swindle of transaction rank;

Bad user's detecting step, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.

According to another aspect of the present invention, also provide the detection system of bad user in a kind of network trading, it is characterized in that, described system comprises:

Transaction swindling detecting unit, for based at least one evidence, the active period of at least one trading object being detected, determines the trading object that has the swindle of transaction rank;

Bad user's detecting unit, carries out at least one in the active period for detection of the trading object of swindling in described existence transaction rank and hands over easy-operating bad user.

Method and system of the present invention can detect the bad user of fraud in the transaction of possibility participation network, and the real information that obtains commodity or service for all-network transaction user provides important foundation, has improved the security of network trading.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of bad user's detection method in network trading in the specific embodiment of the invention;

Fig. 2 a is an example enlivening event in trading object ranking list;

Fig. 2 b is an example of active period in trading object ranking list;

Fig. 3 is that of trading object enlivens the schematic diagram in different rank stages in event;

Fig. 4 a is that the rank of the trading object of doubtful existence transaction rank swindle records schematic diagram;

Fig. 4 b is that the rank of an arm's length transaction object records schematic diagram;

Fig. 5 is the system construction drawing of bad user's detection system in network trading in the specific embodiment of the invention;

Fig. 6 is the structural representation of bad user's detection system in network trading in another embodiment of the present invention.

Embodiment

Below in conjunction with drawings and Examples, the specific embodiment of the present invention is described in further detail.Following examples are used for illustrating the present invention, but are not used for limiting the scope of the invention.

In the present invention, " trading object " should be interpreted broadly, it comprises extensive stock or the service that can be published on network and can browse, inquire about, buy, evaluate for user, both comprised traditional physical commodity, such as household electrical appliance, paper book etc., also comprise virtual goods, such as application program, e-book etc., also comprise various service content, such as network membership service etc.

In the present invention, " bad user " refers to the network user who utilizes fraud in the likely participation network transaction that detection method of the present invention or system identification go out.

Analysis and the research of globality have been carried out in the impact that behavior on bad user in network trading in the specific embodiment of the present invention and these behaviors cause the rank of trading object, provide a kind of can Sampling network bad user's technology in transaction, it can detect by the analysis of the historical information to trading object " active period " of some trading objects, special characteristic for trading object in active period (comprises rank feature, user's evaluating characteristic, user comment feature etc.), based at least one evidence, detect the part trading object that has the swindle of transaction rank, and hand over easy-operating user to detect as bad user by occurring in the active period of these trading objects.

According to inventor's analysis, find, exist the trading object of transaction rank swindle can't in billboard, occupy for a long time very high rank, the situation that rank is higher is only to concentrate and occur in one relatively short period as some independent events, and this rank fraud that shows to conclude the business occurs in this period just.In the present invention, trading object can be continued to " enlivening event (Leading Event) " that rank is called trading object higher period, can be called to " active period (the Leading Session) " of trading object the period of frequently enlivening event.Therefore, for bad user's detection, first need to detect that each trading object likely exists the swindle of transaction rank that this enlivens event and this active period.

Online store operator place may have the various historical informations of trading object, from online store, operator directly obtains, or analyze and process by the trading object ranking list relevant information that online store operator is continued within one period of longer period of history to issue, also can obtain the historical information of trading object.Because this historical information of trading object has been recorded the historical information evaluated about the historical information of trading object rank, about the user of trading object and about much informations such as the historical informations of the user comment of trading object, therefore in the specific embodiment of the invention, can carry out the event of enlivening of each trading object and the detection of active period based on this historical information, and and then the detection of realization to bad user.By the rank behavior of analyzing trading object, find, than normal trading object, the trading object that has a transaction rank swindle can be ready-made different special characteristic enlivening in event and active period.Therefore, likely from the historical information of trading object, extract some for judging the evidence of transaction rank swindle, and obtain these evidences, thereby realize the detection of the trading object to there is the swindle of transaction rank, and finally realize the detection to bad user.

As shown in Figure 1, provide the detection method of bad user in a kind of network trading in a specific embodiment of the present invention, described method comprises:

Active period detecting step S10, detects the active period of at least one trading object based on historical information;

Transaction swindling detecting step S20, detects the active period of described at least one trading object based at least one evidence, determines the trading object that has the swindle of transaction rank;

Bad user's detecting step S30, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.

Each steps flow chart and the function of above-mentioned bad user's detection method in the specific embodiment of the invention are described below, by reference to the accompanying drawings.

Because historical information is to detect bad user's significant data basis in the present invention, therefore as a preferred embodiment of the present invention, this bad user's detection method also can comprise a historical information obtaining step, obtains the historical information of trading object.

Trading object ranking list can show the trading object of K position before welcome rank conventionally, such as first 1000 etc.And trading object ranking list is understood regular update conventionally, upgrade for example every day.Therefore, have its historical ranking information for each trading object a, this history ranking information can comprise and is expressed as a rank sequence corresponding with discrete-time series

interval between time point in this discrete-time series is fixed, i.e. the update cycle of trading object ranking list.Wherein,

that this trading object a is at time t _itime rank,

+ ∞ represents the not row of K position before ranking list rank of trading object a; N represents the corresponding time point sum of all historical ranking informations.For example, in ranking list every day more under news, t _ijust represent the i days in this phase of history, be exactly total the corresponding number of days of the historical ranking information of n.Can find out,

value less, illustrate that the trading object a i days rank in ranking list is higher.

After a trading object is published on network, any network user can evaluate it.In fact, user's evaluation is one of most important feature for trading object is promoted.The trading object with more high praise will attract more users pay close attention to or buy it, and causes the higher rank of this trading object in ranking list.Thereby in historical information, can comprise history evaluation information, i.e. the evaluation information that in historical each time period, the network user makes this trading object.

Equally, after a trading object is published on network, any network user can carry out to it comment of character property.In fact, user comment is one of most important feature for trading object is promoted.The trading object with more actively comment will attract more users pay close attention to or buy it, and causes the higher rank of this trading object in ranking list.Thereby in historical information, can comprise historical review information, i.e. the review information that in historical each time period, the user of trading object makes this trading object.

In this historical information obtaining step, can obtain in many ways this historical information.For example, can directly obtain this historical information from online store operator, the data that also can continue within one period of longer period of history from online store to issue, extract this historical information etc.

S10: active period detecting step, detects the active period of at least one trading object based on historical information.

Active period represents that trading object rank in trading object ranking list is higher, and in namely one higher period of user's attention rate, the transaction rank fraud that therefore can affect greatly network trading only there will be in these active period.So in the specific embodiment of the invention, first will detect the active period of at least one trading object for bad user's detection from the historical information of trading object.

Meanwhile, if but a trading object exists the swindle of transaction rank do not make this trading object occupy the higher ranked in trading object ranking list, this transaction rank fraud can't affect greatly network trading.Therefore,, in active period detecting step, only need to focus on the part trading object that in trading object ranking list, rank is higher, and detect the active period of these trading objects.As for the concrete quantity of detected trading object, can determine a predetermined quantity according to those skilled in the art's demand and processing power etc., the span of this predetermined quantity for example can be between 1～500.

In a preferred embodiment of the invention, in this active period detecting step, can further comprise and enliven event detection step, based on this historical information, detect the event of enlivening of each trading object in the trading object of above-mentioned predetermined quantity.

Because the seller of trading object all wishes its trading object, occupy higher rank in ranking list, therefore trading object seller likely utilizes the means of transaction rank swindle to make its trading object rank among ranking list prostatitis.By analysis, find, trading object can't always occupy very high rank in billboard, there is to continue rank and be " enlivening event " higher period, the example of the event of enlivening of trading object has been shown in Fig. 2 a, in figure, transverse axis represents time series (Date Index) corresponding to historical ranking information in historical information, the longitudinal axis represents the rank (Ranking) of trading object, event 1(Event1 in figure) and event 2(Event2) represent to occur in this trading object placement history two enliven event, its profile is formed by connecting by the rank point enlivening during event respectively.

In the specific embodiment of the invention, the trading object standard that rank is higher in trading object ranking list is that the rank of this trading object is not more than a rank threshold k *.Due to the row of rank K* position before ranking list of trading object, to be considered to rank higher, thereby the rank of trading object continues can be considered to one in the time period of the row of front K* position and enlivens event, this enlivens event and should from this trading object starts to enter ranking list, start by the row of K* position, lasts till that this trading object falls the row end that K* position before ranking list.

Preferably, the method in embodiment of the present invention also can comprise that one arranges the step of this rank threshold k *, thereby determines trading object higher standard of rank in trading object ranking list.Because the trading object total quantity K in ranking list is conventionally very large, be for example 1000 etc., therefore above-mentioned rank threshold k * is less than K value conventionally.According to factors such as the total quantity K of trading object in trading object ranking list and those skilled in the art's analysis demands, this rank threshold k * can be in for example value between the integer between 1～500.The value that it will be understood by those skilled in the art that K* is less, and it is just higher that trading object is considered to the standard that rank is higher.In Fig. 2 a, the value of this K* is 300.

According to the above-mentioned character express for enlivening event, the event of the enlivening e of trading object a formulism statement as follows:

A given rank threshold k * is as the higher standard of rank, wherein K* ∈ [1, K]; The event of the enlivening e of trading object a comprises the time range of time a to end time from the beginning

the rank of corresponding trading object a meets

and $r_{\mathrm{end}}^a\&le;K*<r_{\mathrm{end}+1}^a,$ And $\&ForAll;t_k\&Element;(t_{\mathrm{start}}^e,t_{\mathrm{end}}^e)$ All meet $r_k^a\&le;K*.$

According to above-mentioned statement, can find out, the rank that detects trading object for important being of the detection that enlivens event continued in start time and the end time of a period of time of the row of front K* position, and will be defined as enlivening event the period between a pair of start time and end time.Therefore,, in the specific embodiment of the invention, this enlivens event detection step and can further comprise the steps:

Start time identification step S101: in this step, identify the start time of the event of enlivening in the historical ranking information from historical information.Particularly, in this start time identification step, trading object rank in can sequential search historical ranking information on each time point, when the rank that is not more than rank threshold k * and a upper time point when the rank of current point in time is greater than rank threshold k *, identification current point in time is the start time of enlivening event.It will be understood by those skilled in the art that owing to may comprise multiple events of enlivening in trading object placement history, therefore in this start time identification step, may identify multiple start time points.

End time identification step S102: in this step, identify the end time of the time of enlivening from historical ranking information.Particularly, in this end time identification step, trading object rank in can sequential search historical ranking information on each time point, when the rank that is greater than rank threshold k * and a upper time point when the rank of current point in time is not more than rank threshold k *, identifying a upper time point is the end time of enlivening event.It will be understood by those skilled in the art that owing to may comprise multiple events of enlivening in trading object placement history, therefore in this end time identification step, may identify multiple end time points.

Enliven event recognition step S103: in this step, the time period between each start time and its afterwards adjacent end time is identified as to the event of enlivening, has so just detected all the enliven events of trading object in placement history.

What deserves to be explained is, as a kind of special circumstances, if on first time point of period of history of analyzing and processing, the for example first day in historical record, the rank of trading object is the row of K* position before ranking list just, now, in described start time identification step S101, this first time point is defined as to a start time.Similarly, if on last time point of period of history of analyzing and processing, for example today, the rank of trading object is the row of K* position before ranking list still, now in described end time identification step S102, this last time point are defined as to an end time.

Introduced the mode that detects the event of enlivening of trading object above, on this basis, in a preferred embodiment of the invention, can in this active period detecting step, merge the event of enlivening adjoining in the object of each transaction to form the described active period of this trading object.

By further research, find, can there is repeatedly continuously the near event of enlivening adjacent one another are in some trading objects within one period, and be exactly " active period " of trading object in the present invention this period.Visible, the adjoining event merge that enlivens is got up just to have formed active period.Particularly, can be less than to an interval threshold φ adjacent two time intervals of enlivening event as enlivening the standard of event merge in same active period by two, adjacent two time intervals of enlivening event refer to that adjacent two are enlivened last end time and rear one of enlivening event in event and enliven the interval between start time of event.

Preferably, the method in embodiment of the present invention also can comprise that one arranges the step of this interval threshold φ, thereby determines and enliven the standard of event merge in same active period by two.According to factors such as those skilled in the art's analysis demands, the value of this interval threshold φ can be the round values in 2～10 times of update cycle of trading object ranking list.The value that it will be understood by those skilled in the art that interval threshold φ is less, enlivens the standard of event merge in same active period just higher by two.

The example of the active period of trading object has been shown in Fig. 2 b, in figure, transverse axis represents the time series that historical ranking information is corresponding (Date Index), the longitudinal axis represents the rank (Ranking) of trading object, 1(Session1 during in figure) and during 2(Session2) represent two active period that occur in this trading object placement history, each active period consists of multiple events of enlivening.

According to the above-mentioned character express for active period, the active period s of trading object a formulism statement as follows:

The active period s of trading object a comprises a time range

with n the adjacent event of enlivening { e ₁..., e _n, it meets

and do not exist other active period s* to make $T_s\&SubsetEqual;T_{s*}.$ In addition, $\&ForAll;i\&Element;[1,n)$ Have $(t_{\mathrm{start}}^{e_{i+1}}-e_{\mathrm{end}}^{e_i})<\mathrm{\&phi;},$ Wherein φ be preset enliven interval of events threshold value, be for judging between the event of enlivening that adjacent degree is to include them in the criterion of same active period.

According to above-mentioned statement, can find out, for important being of the detection of active period, based on interval threshold φ, by adjoining in trading object placement history, enliven event merge to form active period.Particularly, in the active period detecting step of the specific embodiment of the invention, initial time point from historical ranking information starts the each event of enlivening detecting of sequential search, when current, enliven event and upper one time interval of enlivening event while being less than this interval threshold φ, enliven event merge in same active period by these two, until searched for all events of enlivening that detect to detect all active period of this trading object in placement history.

What deserves to be explained is, as a kind of special circumstances, if one is enlivened event, to enliven event not adjoining with any other, and this enlivens event self also can be considered to form an active period.In this case, in this active period detecting step, enliven event and upper one time interval of enlivening event was not less than described interval threshold φ when one, and this enlivens event and next and enlivens the time interval of event while being not less than described interval threshold φ, detect this and enliven event from as an active period.

Just as mentioned before, detected above-mentioned active period represents that trading object rank in trading object ranking list is higher, namely be subject to one period that user welcomes, this detected active period can be used as including the data basis of detecting the various trading object services bad user.Therefore,, after detecting the active period of trading object, as a preferred embodiment of the invention, the active period information of detected trading object can also be sent to online store operator or the network user.

For online store operator, it can further analyze the transaction rank fraud that utilizes fraudulent mean to obtain false high rank in ranking list etc. according to this active period information, thereby improves the operation of online store; And for the terminal user of trading object, they can judge that trading object exists the possibility of transaction rank swindle or selection to meet the trading object etc. of self-demand voluntarily according to this active period information.

In addition,, as detecting the event of enlivening of trading object and a kind of specific implementation of active period, following algorithm 1 shows the example of a program code that detects active period in the historical information of given trading object a.

In above-mentioned algorithm 1, each event e that enlivens is defined as

active period s is defined as

wherein E _sit is the set that enlivens event in active period s.Especially, each that first extracts trading object a from the start time of historical ranking information enlivens the step 2-5 event e(algorithm 1).For the each event of enlivening e extracting, detect e and previously enliven time interval between event e* to judge whether they belong to same active period.Particularly, if the event e of enlivening is considered to belong to a new active period (the step 7-13 in algorithm 1).Like this, above-mentioned algorithm 1 can be identified the event of enlivening and active period by the single pass of the historical ranking information to trading object a.

Transaction swindling detecting step S20, detects the active period of described at least one trading object based at least one evidence, determines the trading object that has the swindle of transaction rank.

As a preferred embodiment of the present invention, this transaction swindling detecting step can further comprise an evidence verification step, based at least one evidence, the described active period of the each trading object detecting is verified and obtained a swindle parameter of corresponding trading object.Like this, after extracting particular evidence, can calculate the swindle parameter corresponding with this evidence, this swindle parameter itself can be used as the transaction rank fraud detection result of the transaction swindling detecting step in present embodiment, thereby judges which trading object exists transaction rank swindle phenomenon.Comparatively complicated owing to affecting the factor of special characteristic of trading object, only rely on one or a class evidence possibly cannot accurately judge whether a trading object exists the swindle of transaction rank but only obtain a detected value for reference (swindle parameter), but those skilled in the art can judge that trading object exists the possibility of transaction rank swindle according to this swindle parameter.

In the specific embodiment of the invention, can extract respectively three classes for transaction rank swindle to the evidence detecting, respectively: the evidence relevant to rank, and user evaluates relevant evidence and the relevant evidence with user comment.Just carry out to introduce respectively this three classes evidence below and utilize their conclude the business concrete steps of rank fraud detection in the specific embodiment of the invention.

(1) evidence relevant to rank

As the above introduction to the historical ranking information in historical information, it comprises and is expressed as a rank sequence corresponding with discrete-time series, each element in this rank sequence, corresponding to a discrete time point in time series, represents the rank of this trading object when this discrete time point.Meanwhile, active period is that trading object is likely concluded the business period of rank swindle.Therefore, can analyze the rank feature of historical ranking information in trading object active period, extract some information relevant to rank as the evidence for detection of the swindle of transaction rank.

Because an active period may comprise one or more events of enlivening, therefore in order to extract the evidence for detection of the swindle of transaction rank in active period, as a preferred embodiment of the present invention, this transaction swindling detecting step can further comprise and enlivens event analysis step, analyze each some basic rank features of enlivening event in active period, ascent stage, maintenance stage and decline stage that for example identification enlivens event.

Particularly, known by the historical ranking information of analysis trading object, trading object meets specific rank feature conventionally in the rank behavior enlivening in event, includes three different rank stages: ascent stage, maintenance stage and decline stage.Each, enliven in event, first the rank of trading object rises within the scope of a peak value of ranking list (is ascent stage, Raising Phase), then within the scope of this peak value, keep (keep the stage one period, Maintaining Phase), last rank declines until enliven the end (being the decline stage, Recession Phase) of event.Fig. 3 shows one and enlivens the example in different rank stages in event, and in figure, transverse axis represents the time series that historical ranking information is corresponding (Date Index), and the longitudinal axis represents the rank (Ranking) of trading object.

Based on above-mentioned character express, the above-mentioned three phases that enlivens event is carried out to formulistic description below:

For given trading object a, at it, enliven the time range of event e

in, the highest rank position of trading object a is

it belongs within the scope of Δ R.The ascent stage that enlivens event e refers to time range

wherein $t_a^e=t_{\mathrm{start}}^e,r_b^a\&Element;\mathrm{\&Delta;R}$ And ${\&ForAll;t}_i\text{\&Element;}{[t}_a^e,t_b^e)$ Meet

the maintenance stage of enlivening event e refers to time range, wherein and

meet

the decline stage of enlivening event refers to event scope

wherein $t_d^e=t_{\mathrm{end}}^e.$

It should be noted that in the foregoing description, Δ R determines the start time in maintenance stage and the rank scope of end time,

with

it is respectively the rank of trading object a first time and last time in rank range delta R.Thereby the scope that those skilled in the art can arrange Δ R according to analysis demand is carried out stage division to enlivening event, for example in Fig. 3, the scope of Δ R is trading object rank in ranking list first 70.In a preferred embodiment of the invention, at this, enliven and in event analysis step, identify above-mentioned triphasic mode and be: the rank of determining trading object in enlivening event first time and last time in peak value range delta R, time period between this first time and this last time is identified as to the maintenance stage, by enlivening the time period before the maintenance stage in event, be identified as ascent stage, by enlivening the time period after the maintenance stage in event, be identified as the decline stage.

For a trading object, even if exist transaction rank swindle also can not always remain on an identical peak, for example, always in ranking list, rank the first, but remain within the scope of a peak value, for example, first 25 of ranking list etc.If active period s of trading object a exists the swindle of transaction rank, its active period of enlivening rank behavior meeting in this three phases of event and those normal trading objects is different.In fact, the trading object of each existence transaction rank swindle always has the rank target of an expectation, for example within first 25 of ranking list, keep one week etc., simultaneously also can be according to this rank target for example, to the bad user of employ to implement rank fraud pay (remain in time of first 25 every day 1000 dollars etc.).Therefore,, no matter for the seller of trading object or for the bad user who is employed, they just can make a profit sooner to reach sooner this rank target.In addition, reaching and keeping after one section of required time of this rank target, transaction rank fraud meeting stops, and the rank of this trading object will there will be rapid drawdown.As can be seen here, exist the event of enlivening of transaction rank swindle will present very short ascent stage and very short decline stage.Simultaneously, because swindle the expense that makes trading object occupy a ranking list high position be very high by transaction rank, therefore exist trading object that transaction rank swindles conventionally each, only to have a shorter maintenance stage to make this trading object in a ranking list high position in enlivening event.

Fig. 4 a shows the rank record of the trading object of doubtful existence transaction rank swindle.In the drawings, can find out that this trading object exists the event of enlivening of multiple pulseds.On the contrary, for normal trading object, its rank behavior that enlivens in event is distinct.For example, Fig. 4 b shows the rank record of an arm's length transaction object very popular with users, and it comprises that has the event of enlivening of scope (being longer than 1 year) for a long time, especially in the decline stage.In fact, the higher rank of ranking list once an arm's length transaction object rises, it conventionally has large numbers of buyers and may attract increasing user to go to buy, and therefore this trading object will occupy higher rank for a long time in ranking list.Based on above-mentioned analysis, the present invention can extract some distinguishing marks relevant with rank and build evidence (with rank relevant evidence) in the active period of trading object, and utilizes these evidences to detect the existence of transaction rank swindle.

According to above-mentioned known to enlivening the analysis of event three phases, exist the event of enlivening of transaction rank swindle will present very short ascent stage and very short decline stage, therefore in a preferred implementation, the evidence relevant to rank can some the rank features that ascent stage in event and/or decline stage embody of enlivening based in active period form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging the swindle of transaction rank.

For example, owing to having identified each ascent stage and decline stage of enlivening event in active period in event analysis step enlivening, therefore (as comprised in active period, 3 are enlivened event can to calculate the mean value of the time range of all ascent stages that enliven event in active period, this mean value be the time range of 33 ascent stages that enliven event and again divided by 3), or the mean value of the time range of all decline stages of enlivening event, or the time range of all ascent stages that enliven event and the time range of decline stage and mean value, as this swindle parameter.

Again for example, can also calculate the mean value of the angle of curve the formed acute angle crossing with time shaft of all ascent stages that enliven event in active period, or the mean value of the angle of the curve of all decline stages of enlivening event formed acute angle crossing with time shaft, or the angle of the curve of all ascent stages that enliven event and the curve of decline stage formed acute angle crossing with time shaft and mean value, as this swindle parameter.As shown in Figure 3, two acute angle parameter θ ₁and θ ₂show respectively of trading object a and enliven ascent stage curve in event e (in ascent stage each adjacent rank value point be connected the curve forming) and decline stage curve (in the decline stage each adjacent rank value point be connected the curve of formation) formed acute angle crossing with time shaft.According to enlivening before in event analysis step, for the formulism that enlivens three phases in event, describe, those skilled in the art can calculate above-mentioned parameter θ by following formula ₁and θ ₂:

${\mathrm{\&theta;}}_1^e=\arctan \left(\frac{K*-r_b^a}{t_b^e-t_a^e}\right),{\mathrm{\&theta;}}_2^e=\arctan \left(\frac{K*-r_c^a}{t_d^e-t_c^e}\right)---\left(1\right)$

Wherein K* is the rank threshold value that represents higher ranked.

Can find out θ ₁be worth greatlyr, just represent the trading object a rank within a short period of time higher ranked that rises sharply; θ ₂be worth greatlyr, represent that trading object a is plummeted to rank bottom from higher ranked within very short time.Therefore,, for an active period, if comprising, it there is larger θ more ₁value or larger θ ₂the event of enlivening of value, just shows that it exists the possibility of transaction rank swindle larger.For example, when the angle of the curve of all ascent stages that enliven event in active period and the curve of decline stage formed acute angle crossing with time shaft and mean value during as this swindle parameter, can further describe this swindle parameter here

as follows:

$\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_s}=\frac{1}{\left|E_s\right|}\underset{e\&Element;s}{\mathrm{\&Sigma;}}({\mathrm{\&theta;}}_1^e+{\mathrm{\&theta;}}_2^e)---\left(2\right)$

Wherein | E _s| be the total number of the event of enlivening that comprises in active period s.Visible, than the active period of other trading objects in ranking list, obviously larger if the active period s of a trading object comprises

value, with regard to there is a strong possibility there is the swindle of transaction rank in property to this trading object.

According to above-mentioned known to enlivening the analysis of event three phases, the trading object that has a rank swindle only has a shorter maintenance stage to make this trading object in a ranking list high position each in enlivening event conventionally, therefore in a preferred embodiment of the invention, some rank features that the evidence relevant to rank can embody in the maintenance stage in event of enlivening based in active period form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging the swindle of transaction rank.

For example, owing to having identified each maintenance stage of enlivening event in active period enlivening in event analysis step, therefore can calculate the mean value of the time range in all maintenance stages of enlivening event in active period as this swindle parameter.

Again for example, can average rank and the described time range of enlivening event based on this trading object in all maintenance stages of enlivening event in active period calculate this swindle parameter.Particularly, as discussed above, the trading object that has a transaction rank swindle has the shorter maintenance stage enlivening conventionally in event.Therefore, if used represent the time range in the maintenance stage of enlivening event e, and the average list of trading object a in this maintenance stage is shown

can for example define the swindle parameter Χ of an active period _sas follows:

$X_s=\frac{1}{\left|E_s\right|}\underset{e\&Element;s}{\mathrm{\&Sigma;}}\frac{K*-{\stackrel{\&OverBar;}{r}}_m^e}{\mathrm{\&Delta;}{t}_m^e}---\left(3\right)$

Wherein K* is the rank threshold value that represents higher ranked.Visible, than the active period of other trading objects in ranking list, if the active period s of a trading object comprises obviously larger Χ _svalue, with regard to there is a strong possibility there is the swindle of transaction rank in property to this trading object.

In addition it will be understood by those skilled in the art that, the number of the event of enlivening comprising in the active period s of trading object | E _s| be also the important symbol that has rank swindle.For normal trading object, the decline stage shows the reduction of degree of welcome, therefore after the end that enlivens event, unlikely again occurs in a short time that another enlivens event, unless this trading object has been taked other trade promotion means.Therefore,, than the active period of other trading objects in ranking list, if an active period of trading object has comprised many enliven event more than the active period of other trading objects in ranking list, just there is a strong possibility that property exists the swindle of transaction rank for this trading object.

According to above-mentioned to enlivening event number goal analysis in active period, in a preferred implementation, the evidence relevant to rank can be based in active period the quantity of the event of enlivening form, and this evidence based on formed is determined the quantity of the event of enlivening in active period | E _s|, as the swindle parameter for judging the swindle of transaction rank.

(2) evaluate relevant evidence to user

The evidence relevant with rank is extremely important for detecting the swindle of transaction rank, but sometimes, adopts the evidence relevant with rank always ineffective.For example, some trading object is to be released by famous seller, is subject to the impact of seller's prestige and public praise, and the ascent stage of the event of enlivening of these trading objects has very large θ ₁value.In addition, be subject to the impact of the legal market service such as some for example " in limited time discounts ", also can cause the appearance of some evidences relevant with rank.In order to address these problems, in the specific embodiment of the invention, how research is simultaneously extracted other features and is used as detecting the evidence that transaction rank is swindled from historical information.

As the above introduction to historical information, it comprises history evaluation information, i.e. the user that in historical each time period, the user of trading object makes this trading object evaluates.Meanwhile, active period is that trading object is likely concluded the business period of rank swindle.Therefore, can analyze the evaluating characteristic of history evaluation information in trading object active period, extract some and evaluate relevant information as the evidence for detection of the swindle of transaction rank to user.

Particularly, after a trading object is published, any network user can evaluate it, for example, this trading object is provided to the scoring of 1～5 point, common 5 points of representative of consumer are satisfied with (high praise) to this trading object very much, and 1 point of very dissatisfied (minimum evaluation) of representative.In fact, user's evaluation is one of most important feature for trading object is promoted.The trading object with more high praise will attract more users to buy it, and causes the higher rank of this trading object in ranking list.Therefore, false evaluation is also the important behaviour form in the swindle of transaction rank.If there is the swindle of transaction rank in the active period s of trading object, evaluation within the time period of active period s will have the off-note different from the evaluation of other historical stages, and this feature can be used for building evaluates relevant evidence for detection of transaction rank swindle to user.

For normal trading object, the average user evaluation within the particular active phase should be consistent with the average ratings in its all history evaluation records.On the contrary, for the trading object that has the swindle of transaction rank, in its active period, than its history evaluation, can there is surprising high praise.As a preferred embodiment of the present invention, evaluate to user the average user evaluation that relevant evidence can be based in active period

with historical average ratings

form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging the swindle of transaction rank.

For example intuitively, can calculate the mean value that in active period, all users evaluate

with historical average ratings

between difference, or all users evaluate mean value

with historical average ratings

between ratio, as this swindle parameter.

Again for example, can also calculate the mean value that in active period, all users evaluate

with historical average ratings

between difference and historical average ratings

ratio, as this swindle parameter.This swindle parameter Δ R is described by formulism _sas follows:

$\mathrm{\&Delta;}{R}_s=\frac{{\stackrel{\&OverBar;}{R}}_s-{\stackrel{\&OverBar;}{R}}_a}{{\stackrel{\&OverBar;}{R}}_a},(s\&Element;a)---\left(4\right)$

Wherein

the average user evaluation of estimate in active period,

the history evaluation mean value of trading object a.Therefore, than the active period of other trading objects in ranking list, if the active period s of a trading object comprises obviously larger Δ R _svalue, with regard to there is a strong possibility there is the swindle of transaction rank in property to this trading object.

In the evaluation information of trading object, each evaluation can be classified as a discrete opinion rating system | in L|, for example, comprise that it has represented the fancy grade of user for this trading object from 1～5 these five grades.For a normal trading object a, its opinion rating l in active period s _idistribution p (l _i| R _s,a) should with its history evaluation record in distribution p (l _i| R _a) be consistent.As a preferred embodiment of the present invention, to user evaluate relevant evidence can be based on trading object the opinion rating in active period distribute and history evaluation information in opinion rating distribute and form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging that transaction rank is swindled.

For example, can calculate the difference between the distribution of opinion rating in the distribution of the opinion rating of trading object in active period and history evaluation information, as this swindle parameter.Particularly, first can pass through

calculate p (l _i| R _s,a) value, wherein

to be l in active period inner evaluation grade _iuser evaluate number,

it is evaluation number total in active period s; Can calculate p (l by similar mode simultaneously _i| R _a); Then calculate the difference between the distribution of opinion rating in the distribution of the opinion rating of trading object in active period and history evaluation information.As a kind of specific implementation, can use p (l _i| R _s,a) and p (l _i| R _a) between cosine distance B (s) estimate the difference between them.By formulism, describe, this swindle parameter D (s) is as follows:

$D\left(s\right)=\frac{{\mathrm{\&Sigma;}}_{i=1}^{\left|L\right|}p\left(l_i\right|R_{s,a})\&times;p\left(l_i\right|R_a)}{\sqrt{{\mathrm{\&Sigma;}}_{i=1}^{\left|L\right|}p{\left(l_i\right|R_{s,a})}^2}\&times;\sqrt{{\mathrm{\&Sigma;}}_{i=1}^{\left|L\right|}p{\left(l_i\right|R_a)}^2}}---\left(5\right)$

Visible, than the active period of other trading objects in ranking list, if the active period s of a trading object comprises obviously larger D (s) value, just there is a strong possibility that property exists transaction rank to swindle for this trading object.

(3) evidence relevant with user comment

As the above introduction to historical information, it comprises historical review information, i.e. the user comment that in historical each time period, the user of trading object makes this trading object.Meanwhile, active period is that trading object is likely concluded the business period of rank swindle.Therefore, can analyze the user comment feature of historical information in trading object active period, extract some information relevant to user comment as the evidence for detection of the swindle of transaction rank.

Particularly, after a trading object is published, most online stores allows user for trading object, to write out the user comment of text formatting.These user comments can reflect personal view or the experience of user to particular transaction object.In fact, user comment is one of most important feature for trading object is promoted, and the user comment of simultaneously forging is also that transaction rank is swindled one of most important aspect.Before buying new trading object; user conventionally can first browse user comment in historical review information and help them and make a decision; the trading object with more positive comment will attract more users to buy it, and causes the higher rank of this trading object in ranking list.Therefore, rank fake producer conventionally can be for the user comment of particular transaction object publishing falseness to stimulate the purchase volume of this trading object, thereby promote rapidly the rank of this trading object in ranking list.If there is the swindle of transaction rank in the active period s of trading object, user comment within the time period of active period s will have the off-note different from the user comment of other historical stages, and this feature can be used for building the evidence relevant to user comment for detection of the swindle of transaction rank.

In fact, because human cost is too high, most fictitious users comments are all implemented by predefined machine.Therefore, user comment fake producer is conventionally frequent issues a large amount of identical or similar user comments to promote the rank of this trading object.On the contrary, because different user has different personal views and experience, normal trading object can have multifarious user comment conventionally.As a preferred embodiment of the present invention, the evidence relevant to user comment can form based on the similarity degree between user comment in active period, and this evidence based on formed calculates an evidence value as the swindle parameter for judging the swindle of transaction rank.

For example, can calculate in active period s the average similarity Sim (s) between user comment as this swindle parameter.Particularly, can following steps calculate this swindle parameter S im (s):

First, each the user comment c in active period s is carried out to standardization.For example, for Chinese user comment, can by " ", the function word such as " this " deletes, for English user comment, the words such as " of ", " the " can be deleted, and for example, by (plays are become to play, better is become to good etc.) such as verb, adjectival distortion removals.

Then, for each user comment c builds standardization vocabulary vector wherein n is the total quantity of all different standardization vocabulary in the interior all user comments of active period s.Particularly, can have

wherein freq _i,cthe frequency of occurrences of i vocabulary in user comment c.

Finally, can pass through cosine similarity

calculate user comment c _iwith user comment c _jbetween similarity.Therefore, can calculate swindle parameter S im (s) by for example following formula:

$\mathrm{Sim}\left(s\right)=\frac{2\&times;{\mathrm{\&Sigma;}}_{1\&le;i\&le;j\&le;N_s}\mathrm{Cos}(\stackrel{\&RightArrow;}{{\mathrm{\&omega;}}_{c_i}},\stackrel{\&RightArrow;}{{\mathrm{\&omega;}}_{c_j}})}{N_s\&times;(N_s-1)}---\left(6\right)$

Wherein N _sit is the total number of user comment in active period s.

Visible, Sim (s) value more just illustrates and in active period s, comprises how same or analogous user comment.Therefore,, than the active period of other trading objects in ranking list, if the active period s of a trading object comprises obviously larger Sim (s) value, just there is a strong possibility that property exists transaction rank to swindle for this trading object.

By the user comment of trading object is analyzed to discovery, each user comment c can be relevant to a specific potential theme z.For example, some user comments are relevant to potential theme " worth purchase ", and some user comments are relevant at theme " poor quality " to label.Meanwhile, because different user can have different individual preferences to trading object, each trading object a should have different themes and distribute in its user comment historical record.For a normal trading object a, the theme distribution p (z|s) of the user comment in its active period s should be consistent with the theme distribution p (z|a) of this trading object a user comment in whole historical record.On the contrary, if a trading object exists fictitious users comment in its active period s, above-mentioned two kinds of themes distribute and will there will be notable difference, for example, in active period, there will be more active user comment, as " being worth buying ", " welcome " etc.As a preferred embodiment of the present invention, the evidence relevant to user comment can based on the theme of trading object user comment in active period distribute and historical review information in the theme of user comment distribute and form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging that transaction rank is swindled.

For example, can calculate trading object user comment in active period theme distribute and historical review information in user comment theme distribute between difference, as this swindle parameter.

Exist in the prior art various for extracting the theme modeling technique of potential theme.In the specific embodiment of the invention, can adopt the potential Di Likelei apportion model (Latent Dirichlet Allocation Model) extensively adopting in prior art to extract all potential theme (D.M.Blei in user comment, A.Y.Ng, and M.I.Jordan.Latent dirichlet allocation.Journal of Machine Learning Research, Pages993-1022,2003).Afterwards, the theme that all potential theme in can the user comment based on extracted calculates trading object user comment in active period distribute and historical review information in the theme of the user comment difference between distributing.

Particularly, first can pass through

calculate p (z _i| value s), wherein

be in active period s user comment theme as z _iuser comment number,

it is user comment number total in active period s; Can calculate p (z by similar mode simultaneously _i| a); Then calculate trading object user comment in active period theme distribute and historical review information in user comment theme distribute between difference.As a kind of specific implementation, can use p (z _i| s) with p (z _i| the cosine distance B (s) a) is estimated the difference between them.By formulism, describe, this swindle parameter D (s) is as follows:

$D\left(s\right)=\frac{{\mathrm{\&Sigma;}}_{i=1}^{M}p\left(z_i\right|s)\&times;p\left(z_i\right|a)}{\sqrt{{\mathrm{\&Sigma;}}_{i=1}^M{p\left(z_i\right|s)}^2}\&times;\sqrt{{\mathrm{\&Sigma;}}_{i=1}^{M}p{\left(z_i\right|a)}^2}}---\left(7\right)$

Wherein, M is the total quantity of the theme of extracted user comment.Visible, than the active period of other trading objects in ranking list, if the active period s of a trading object comprises obviously larger D (s) value, just there is a strong possibility that property exists transaction rank to swindle for this trading object.

Introduced the multiple evidence in three classes and every class above, except the detection of swindling by one in them rank of concluding the business separately in above-mentioned each preferred implementation, in a preferred implementation of evidence verification step, can also consider multiple in above-mentioned evidence, to based on these evidences, verify that the correspondence swindle parameter obtaining is weighted, thereby obtain a final swindle parameter.Consider that above-mentioned multiple evidence likely has different dimensions, those skilled in the art can be according to the attention degree for each evidence in actual analysis demand, the weighted value of determining each swindle parameter based on method for normalizing commonly known in the art and Weight Determination, does not repeat them here.

More than introduced the evidence verification step in transaction swindling detecting step, it can verify and obtain based at least one evidence a swindle parameter to described active period, this swindle parameter itself just can be used as the transaction rank fraud detection result of transaction swindling detecting step.But in order to make those skilled in the art's rank fraud detection of concluding the business more easily, in a preferred implementation, transaction swindling detecting step can further include a swindle parameter determining step, the swindle parameter calculating according to evidence and a threshold value are compared, thereby judge intuitively, judge whether trading object exists the swindle of transaction rank.

It will be appreciated by those skilled in the art that, multiple evidence in three classes and every class based on above introducing, those skilled in the art can arrange respectively corresponding threshold value according to the heterogeneity of evidence and detection demand, according to set threshold value, carry out the judgement whether trading object exists transaction rank to swindle, and using the transaction rank fraud detection result of net result transaction swindling detecting step in the specific embodiment of the invention of judgement.For example, for the evidence relevant for the multiple and rank above introduced, if swindle parameter is to enliven the mean value of the ascent stage of event and/or the time range of decline stage, or the mean value of the time range in maintenance stage, when the swindle parameter calculating is less than set threshold value, judges this trading object and have transaction rank swindle phenomenon; And if swindle parameter is other situations about introducing, when the swindle parameter calculating exceedes set threshold value, judges this trading object and have transaction rank swindle phenomenon.Again for example, for the multiple and user that above introduced evaluates relevant evidence, when the swindle parameter calculating exceedes set threshold value, judge this trading object and have transaction rank swindle phenomenon.Again for example, for the evidence relevant for the multiple and user comment above introduced, when the swindle parameter calculating exceedes set threshold value, judge this trading object and have transaction rank swindle phenomenon.

In transaction swindling detecting step, obtain concluding the business after rank fraud detection result, in a preferred embodiment of the invention, obtained transaction rank fraud detection result can also be sent to online store operator or the network user.For online store operator, it can improve according to this transaction rank fraud detection result the operation in trading object shop; And for the terminal user of trading object, they can select according to this transaction rank fraud detection result trading object meeting self-demand etc.

Bad user's detecting step S30, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.

In transaction swindling detecting step, determined the trading object of all existence transaction ranks swindle, and according to analysis before, the swindle of transaction rank all appears in the active period of trading object, as can be seen here, the network user who carries out relationship trading operation in the active period of trading object that has the swindle of transaction rank probably carries out the bad user that this transaction rank is swindled just.

Based on above-mentioned analysis, in this bad user's detecting step, can be based on historical information, inquiry was carried out at least one and was handed over easy-operating user in all active period of All Activity object that have the swindle of transaction rank, using them as the bad user who detects.Corresponding with defined Evidence type in transaction swindling detecting step, said transaction operation here comprises purchase, evaluation and the comment behavior etc. to trading object.

In above-mentioned network trading in bad user's detection method, by the analysis of the historical information to some trading objects, detect the active period of these trading objects, for the special characteristic (comprising rank feature, user's evaluating characteristic, user comment feature etc.) of trading object in active period, based at least one evidence, detect the part trading object that has the swindle of transaction rank, and hand over easy-operating user to detect as bad user by occurring in the active period of these trading objects.The method can detect the bad user of fraud in the transaction of possibility participation network, and the real information that obtains commodity or service for all-network transaction user provides important foundation, has improved the security of network trading; And, testing result of the present invention can be used as the important foundation data of further analysis network trading behavior, for example, can carry out bad user's credit assessment for detected bad user, according to the network user's prestige grading, carry out the credit assessment of trading object, and then the seller of trading object is carried out to credit assessment etc.

As shown in Figure 5, also provide the detection system 100 of bad user in a kind of network trading in embodiment of the present invention, described system 100 comprises:

Active period detecting unit 110, for detecting the active period of at least one trading object based on historical information;

Transaction swindling detecting unit 120, for based at least one evidence, the active period of described at least one trading object being detected, determines the trading object that has the swindle of transaction rank;

Bad user's detecting unit 130, carries out at least one in the active period for detection of the trading object of swindling in described existence transaction rank and hands over easy-operating bad user.

Each Elementary Function of said detecting system is described below, by reference to the accompanying drawings.

Because historical information is to detect bad user's significant data basis in the present invention, therefore as a preferred embodiment of the present invention, this bad user's detection system 100 also can comprise a historical information acquiring unit, for obtaining the historical information of trading object.

This historical information acquiring unit can obtain this historical information in many ways.For example, can directly obtain this historical information from online store operator, the data that also can continue within one period of longer period of history from online store to issue, extract this historical information etc.

Active period detecting unit 110, for detecting the active period of at least one trading object based on historical information.

In a preferred embodiment of the invention, this active period detecting unit 110 can further comprise and enliven event checking module, for detect the event of enlivening of each trading object of at least one trading object based on this historical information.

Preferably, the system in embodiment of the present invention also can comprise a rank threshold value setting unit, for the value of rank threshold k * is set, thereby determines trading object higher standard of rank in trading object ranking list.The value of this rank threshold k * can be the integer between 1～500.

In the specific embodiment of the invention, this enlivens event checking module and further comprises:

Start time identification module 111, for identifying the start time of the event of enlivening from the historical ranking information of historical information.Particularly, trading object rank in can the sequential search historical ranking information of this start time identification module on each time point, when the rank that is not more than rank threshold k * and a upper time point when the rank of current point in time is greater than rank threshold k *, identification current point in time is the start time of enlivening event.

End time identification module 112, for identifying the end time of the time of enlivening from historical ranking information.Particularly, trading object rank in can the sequential search historical ranking information of this end time identification module on each time point, when the rank that is greater than rank threshold k * and a upper time point when the rank of current point in time is not more than rank threshold k *, identifying a upper time point is the end time of enlivening event.

Enliven event recognition module 113, for the time period between each start time and its afterwards adjacent end time is identified as to the event of enlivening, so just detected all the enliven events of trading object in placement history.

What deserves to be explained is, as a kind of special circumstances, if on first time point of period of history of analyzing and processing, the for example first day in historical record, the rank of trading object is the row of K* position before ranking list just, and now this start time identification module 111 is defined as a start time by this first time point.Similarly, if on last time point of period of history of analyzing and processing, for example today, the rank of trading object is the row of K* position before ranking list still, and now this end time identification module 112 is defined as an end time by this last time point.

In a preferred embodiment of the invention, this active period detecting unit 110 is for merging the adjoining event of enlivening to form the described active period of described trading object.

Preferably, the rank fraud detection system 100 in embodiment of the present invention also can comprise an interval threshold setting unit, for the value of this interval threshold φ is set, thereby determines and enlivens the standard of event merge in same active period by two.The value of this interval threshold φ can be the round values in 2～10 times of update cycle of trading object ranking list.

In the specific embodiment of the invention, the initial time point of active period detecting unit 110 from historical ranking information starts the each event of enlivening detecting of sequential search, when current, enliven event and upper one time interval of enlivening event while being less than this interval threshold φ, enliven event merge in same active period by these two, until searched for all events of enlivening that detect to detect all active period of this trading object in placement history.

What deserves to be explained is, as a kind of special circumstances, if one is enlivened event, to enliven event not adjoining with any other, and this enlivens event self also can be considered to form an active period.In this case, this active period detecting unit 110 is for enlivening event and upper one time interval of enlivening event was not less than described interval threshold φ when one, and this enlivens event and next and enlivens the time interval of event while being not less than described interval threshold φ, detect this and enliven event from as an active period.

As a preferred embodiment of the invention, rank fraud detection system 100 can also comprise an active period transmitting element, and the active period information of detected trading object is sent to online store operator or the network user.

Transaction swindling detecting unit 120, for based at least one evidence, the active period of described at least one trading object being detected, determines the trading object that has the swindle of transaction rank.

As a preferred embodiment of the present invention, this transaction swindling detecting unit 120 can further comprise an evidence authentication module, for the described active period of the each trading object detecting being verified based at least one evidence and obtained a swindle parameter of corresponding trading object.

In the specific embodiment of the invention, can extract respectively the evidence relevant to rank, evaluate relevant evidence with user and the relevant evidence with user comment.Just carry out to introduce respectively this transaction swindling detecting unit 120 below in the present invention based on the conclude the business embodiment of rank fraud detection of this three classes evidence.

(1) evidence relevant to rank

Because an active period may comprise one or more events of enlivening, therefore in order to extract the evidence for detection of the swindle of transaction rank in active period, as a preferred embodiment of the present invention, this transaction swindling detecting unit 120 can further comprise and enlivens event analysis module, for example, for analyzing each some basic rank features of enlivening event in active period, identification enlivens event ascent stage, maintenance stage and decline stage.In a preferred embodiment of the invention, at this, enliven the above-mentioned triphasic mode of event analysis module identification and be: the rank of determining trading object in enlivening event first time and last time in peak value range delta R, time period between this first time and this last time is identified as to the maintenance stage, by enlivening the time period before the maintenance stage in event, be identified as ascent stage, by enlivening the time period after the maintenance stage in event, be identified as the decline stage.

In a preferred implementation, the evidence relevant to rank can some the rank features that ascent stage in event and/or decline stage embody of enlivening based in active period form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging the swindle of transaction rank.In another preferred implementation, some rank features that the evidence relevant to rank can embody in the maintenance stage in event of enlivening based in active period form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging the swindle of transaction rank.In another preferred implementation, the quantity of the event of enlivening that the evidence relevant to rank can be based in active period forms, and this evidence based on formed is determined the quantity of the event of enlivening in active period | E _s|, as the swindle parameter for judging the swindle of transaction rank.

(2) evaluate relevant evidence to user

In a preferred implementation, evaluate to user the average user evaluation that relevant evidence can be based in active period

with historical average ratings

form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging the swindle of transaction rank.In another preferred implementation, to user evaluate relevant evidence can be based on trading object the opinion rating in active period distribute and history evaluation information in opinion rating distribute and form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging that transaction rank is swindled.

(3) evidence relevant to user comment

In a preferred implementation, the evidence relevant to user comment can form based on the similarity degree between user comment in active period, and this evidence based on formed calculates an evidence value as the swindle parameter for judging the swindle of transaction rank.In another preferred implementation, the evidence relevant to user comment can based on the theme of trading object user comment in active period distribute and historical review information in the theme of user comment distribute and form, and this evidence based on formed calculates an evidence value as the swindle parameter for judging that transaction rank is swindled.

Except in above-mentioned each preferred implementation separately with above-mentioned all kinds of and all kinds of in the rank fraud detection of concluding the business in various evidences, evidence authentication module can also consider multiple in above-mentioned evidence, to based on these evidences, verify that the correspondence swindle parameter obtaining is weighted, thereby obtain a final swindle parameter.

In order to make those skilled in the art's rank fraud detection of concluding the business more easily, in a preferred implementation, transaction swindling detecting unit 120 can further include a swindle parameter judge module, for the swindle parameter calculating according to evidence and a threshold value are compared, thereby judge intuitively, judge whether trading object exists the swindle of transaction rank.

After the result of rank fraud detection that obtains concluding the business, in a preferred embodiment of the invention, bad user's detection system 100 also comprises a transaction rank fraud detection result transmitting element, and the testing result of obtained transaction rank swindle is sent to online store operator or the network user.

Bad user's detecting unit 130, carries out at least one in the active period for detection of the trading object of swindling in described existence transaction rank and hands over easy-operating bad user.

Particularly, this bad user's detecting unit 130 can be based on historical information, and inquiry was carried out at least one and handed over easy-operating user in all active period of All Activity object that have the swindle of transaction rank, using them all as the bad user who detects.With defined Evidence type is corresponding before, said transaction operation here comprises purchase, evaluation and the comment behavior etc. to trading object.

In above-mentioned network trading, bad user's detection system detects the active period of these trading objects by the analysis of the historical information to some trading objects, for the special characteristic (comprising rank feature, user's evaluating characteristic, user comment feature etc.) of trading object in active period, based at least one evidence, detect the part trading object that has the swindle of transaction rank, and hand over easy-operating user to detect as bad user by occurring in the active period of these trading objects.This system can detect the bad user of fraud in the transaction of possibility participation network, and the real information that obtains commodity or service for all-network transaction user provides important foundation, has improved the security of network trading; And, testing result of the present invention can be used as the important foundation data of further analysis network trading behavior, for example, can carry out bad user's credit assessment for detected bad user, according to the network user's prestige grading, carry out the credit assessment of trading object, and then the seller of trading object is carried out to credit assessment etc.

It will be appreciated by those skilled in the art that, when in the event of enlivening of trading object and the situation of active period ten-four, those skilled in the art can be directly enliven event and active period information is implemented above-mentioned transaction swindling detecting step and bad user's detecting step according to above-mentioned, thereby realize the detection of bad user in network trading.Therefore, bad user's detection method also provide a kind of network trading in another embodiment of the present invention in, described method comprises: transaction swindling detecting step, based at least one evidence, the active period of at least one trading object is detected, determine the trading object that has the swindle of transaction rank; Bad user's detecting step, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.In bad user's detection method of this embodiment, the technology contents details of implementing is identical with bad user's detecting step with transaction swindling detecting step in embodiment before, repeats no more herein.

Accordingly simultaneously, a kind of bad user's detection system of network trading is also provided in another embodiment of the present invention, described system comprises: transaction swindling detecting unit, for based at least one evidence, the active period of at least one trading object being detected, determine the trading object that has the swindle of transaction rank; Bad user's detecting unit, carries out at least one in the active period for detection of the trading object of swindling in described existence transaction rank and hands over easy-operating bad user.In bad user's detection system of this embodiment, the technology contents of implementing is identical with bad user's detecting unit with transaction swindling detecting unit in embodiment before, repeats no more herein.

The structural representation of bad user's detection system 600 in a kind of network trading that Fig. 6 provides for the embodiment of the present invention, the specific embodiment of the invention does not limit the specific implementation of bad user's detection system 600.As shown in Figure 6, this bad user's detection system 600 can comprise:

Processor (processor) 610, communication interface (Communications Interface) 620, storer (memory) 630 and communication bus 640.Wherein:

Processor 610, communication interface 620 and storer 630 complete mutual communication by communication bus 640.

Communication interface 620, for net element communication such as client etc.

Processor 610, for executive routine 632, specifically can realize described in above-mentioned Fig. 5 the correlation function of bad user's detection system in embodiment.

Particularly, program 632 can comprise program code, and described program code comprises computer-managed instruction.

Processor 610 may be a central processor CPU, or specific integrated circuit ASIC(Application Specific Integrated Circuit), or be configured to implement one or more integrated circuit of the embodiment of the present invention.

Storer 630, for depositing program 632.Storer 630 may comprise high-speed RAM storer, also may also comprise nonvolatile memory (non-volatile memory), for example at least one magnetic disk memory.Program 632 specifically can be for realizing following steps:

Active period detecting step, detects the active period of at least one trading object based on historical information;

Transaction swindling detecting step, detects the active period of described at least one trading object based at least one evidence, determines the trading object that has the swindle of transaction rank;

Bad user's detecting step, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.

Program 632 specifically also can be for realizing following steps:

Transaction swindling detecting step, detects the active period of at least one trading object based at least one evidence, determines the trading object that has the swindle of transaction rank;

Bad user's detecting step, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.

In program 632, the specific implementation of each unit can, referring to the corresponding units in each embodiment above, be not repeated herein.

Those skilled in the art can be well understood to, and for convenience and simplicity of description, the specific works process of the equipment of foregoing description and module, can describe with reference to the correspondence in aforementioned means embodiment, does not repeat them here.

Those of ordinary skills can recognize, unit and the method step of each example of describing in conjunction with embodiment disclosed herein, can realize with the combination of electronic hardware or computer software and electronic hardware.These functions are carried out with hardware or software mode actually, depend on application-specific and the design constraint of technical scheme.Professional and technical personnel can realize described function with distinct methods to each specifically should being used for, but this realization should not thought and exceeds scope of the present invention.

If described function realizes and during as production marketing independently or use, can be stored in a computer read/write memory medium using the form of SFU software functional unit.Based on such understanding, the part that technical scheme of the present invention contributes to original technology in essence in other words or the part of this technical scheme can embody with the form of software product, this computer software product is stored in a storage medium, comprise that some instructions (can be personal computers in order to make a computer equipment, server, or the network equipment etc.) carry out all or part of step of method described in each embodiment of the present invention.And aforesaid storage medium comprises: various media that can be program code stored such as USB flash disk, portable hard drive, ROM (read-only memory) (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disc or CDs.

Above embodiment is only for illustrating the present invention; and be not limitation of the present invention; the those of ordinary skill in relevant technologies field; without departing from the spirit and scope of the present invention; can also make a variety of changes and modification; therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.

Claims (74)

1. bad user's a detection method in network trading, is characterized in that, described method comprises:

Active period detecting step, detects the active period of at least one trading object based on historical information;

Transaction swindling detecting step, detects the active period of described at least one trading object based at least one evidence, determines the trading object that has the swindle of transaction rank; And

Bad user's detecting step, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.

2. method according to claim 1, is characterized in that, described active period detecting step further comprises:

Enliven event detection step, based on described historical information, detect the event of enlivening of described at least one trading object;

Active period determining step, merges the adjoining described event of enlivening to form the described active period of described at least one trading object.

3. method according to claim 2, it is characterized in that, the described event of enlivening is that described trading object continues the higher time period of rank in trading object ranking list, and the standard that rank is higher is that the rank of described trading object in trading object ranking list is not more than a rank threshold k *.

4. method according to claim 3, is characterized in that, described method also comprises: described rank threshold k * is set.

5. method according to claim 3, is characterized in that, the span of described rank threshold k * is the integer between 1～500.

6. method according to claim 2, is characterized in that, adjacent two time intervals of enlivening event are less than to an interval threshold φ as enlivening the standard of event merge in same active period by described two.

7. method according to claim 6, is characterized in that, described method also comprises: described interval threshold φ is set.

8. method according to claim 6, is characterized in that, the span of described interval threshold φ is 2～10 times of update cycle of described trading object ranking list.

9. method according to claim 2, is characterized in that, described, enlivens in event detection step, further comprises:

Start time identification step, identifies the described start time of enlivening event;

End time identification step, identifies the described end time of enlivening the time;

Enliven event recognition step, the time period between each start time and its afterwards adjacent end time is identified as to the event of enlivening, thereby detects all events of enlivening of described trading object.

10. method according to claim 9, is characterized in that,

In described start time identification step, the rank of each the above trading object of time point in the historical ranking information of historical information described in sequential search, when the rank that is not more than a rank threshold k * and a upper time point when the rank of current point in time is greater than described rank threshold k *, identification current point in time is the described described start time of enlivening event;

In described end time identification step, the rank of each the above trading object of time point in historical ranking information described in sequential search, when the rank that is greater than described rank threshold k * and a upper time point when the rank of current point in time is not more than described rank threshold k *, identifying a upper time point is the described described end time of enlivening event.

11. methods according to claim 9, is characterized in that,

If the rank of first the above trading object of time point of the historical ranking information of described historical information is not more than described rank threshold k *, in described start time identification step, described first time point of identification is time at the beginning;

If the rank of last the above trading object of time point of described historical ranking information is not more than described rank threshold k *, in described end time identification step, described last time point of identification is an end time.

12. methods according to claim 2, is characterized in that,

In described active period determining step, initial time point from the historical ranking information of described historical information starts the each event of enlivening detecting of sequential search, when current, enliven event and upper one time interval of enlivening event while being less than an interval threshold φ, enliven event merge in same active period by these two, until searched for all events of enlivening that detect to detect all active period of described trading object.

13. methods according to claim 12, it is characterized in that, in described active period determining step, enliven event and upper one time interval of enlivening event was not less than described interval threshold φ when one, and this enlivens event and next and enlivens the time interval of event while being not less than described interval threshold φ, detect this and enliven event from as an active period.

14. methods according to claim 1, is characterized in that, described transaction swindling detecting step further comprises:

Evidence verification step, detects the active period of described at least one trading object based at least one evidence, and obtains a swindle parameter of each trading object.

15. methods according to claim 14, is characterized in that, described transaction swindling detecting step further comprises:

Enliven event analysis step, identify that in described active period, at least one enlivens ascent stage, maintenance stage and the decline stage of event.

16. methods according to claim 15, it is characterized in that, described, enliven in event analysis step, determine the described rank of enlivening trading object described in event first time and last time in a peak value range delta R, time period between described first time and described last time is identified as to the described maintenance stage, by described, enliven the time period before the maintenance stage in event and be identified as described ascent stage, by described, enliven the time period after the maintenance stage in event and be identified as the described decline stage.

17. methods according to claim 15, is characterized in that, ascent stage and/or the decline stage of enlivening in event of described evidence based in described active period forms.

18. methods according to claim 17, is characterized in that,

Described swindle parameter is the mean value of the time range of all described ascent stages that enliven event in described active period, or the mean value of time range of all described decline stages of enlivening event in described active period, or in described active period the time range of all described ascent stages that enliven event and the time range of decline stage and mean value.

19. methods according to claim 17, is characterized in that,

Described swindle parameter is the mean value of the angle of curve the formed acute angle crossing with time shaft of all described ascent stages that enliven event in described active period, or the mean value of the angle of the curve of all described decline stages of enlivening event formed acute angle crossing with time shaft, or the angle of the curve of all described ascent stages that enliven event and the curve of described decline stage formed acute angle crossing with time shaft and mean value.

20. methods according to claim 15, is characterized in that,

The maintenance stage of enlivening in event of described evidence based in described active period forms.

21. methods according to claim 20, is characterized in that,

Described swindle parameter is the mean value of the time range in all described maintenance stages of enlivening event in described active period.

22. methods according to claim 20, is characterized in that, average rank and the time range in described maintenance stage based on trading object described in all described maintenance stages of enlivening event are calculated described swindle parameter.

23. methods according to claim 14, is characterized in that,

The quantity of described evidence based on enlivening event in described active period forms.

24. methods according to claim 23, is characterized in that,

Described swindle parameter is in described active period, to enliven the quantity of event.

25. methods according to claim 14, is characterized in that, average ratings and the historical average ratings of described evidence based in described active period forms.

26. methods according to claim 25, is characterized in that,

Described swindle parameter is average ratings in described active period and difference or the ratio of historical average ratings.

27. methods according to claim 25, is characterized in that,

Described swindle parameter is average ratings and the difference of historical average ratings and the ratio of historical average ratings in described active period.

28. methods according to claim 14, is characterized in that,

In the distribution of described evidence opinion rating in described active period based on trading object and history evaluation information, the distribution of opinion rating forms.

29. methods according to claim 28, is characterized in that,

Described swindle parameter is the difference between the distribution of opinion rating in the distribution of the opinion rating of trading object in described active period and history evaluation information.

30. methods according to claim 29, is characterized in that, the cosine distance in the distribution of the opinion rating by calculating trading object in described active period and history evaluation information between the distribution of opinion rating is calculated the difference between them.

31. methods according to claim 14, is characterized in that, described evidence forms based on the similarity degree between user comment in described active period.

32. methods according to claim 31, is characterized in that,

Described swindle parameter is the average similarity between user comment in described active period.

33. methods according to claim 32, is characterized in that,

Described evidence verification step further comprises:

All user comments in described active period are carried out to standardization;

For each user comment in described active period builds standardization vocabulary vector;

Based on the average similarity between user comment in active period described in described standardization vocabulary vector calculation.

34. methods according to claim 14, is characterized in that,

Described evidence based on the theme of trading object user comment in described active period, distribute and historical review information in the theme of the user comment formation that distributes.

35. methods according to claim 34, is characterized in that,

Described swindle parameter be trading object user comment in described active period theme distribute and historical review information in user comment theme distribute between difference.

36. methods according to claim 35, is characterized in that, by calculate that the theme of trading object user comment in described active period distributes and historical review information in the theme of the user comment cosine distance between distributing calculate the difference between them.

37. methods according to claim 14, it is characterized in that, in described evidence verification step, consider described at least one evidence, to based on described at least one evidence, verify that the correspondence swindle parameter obtaining is weighted, thereby obtain described swindle parameter.

38. according to the method described in any one in claim 14-37, it is characterized in that, described transaction swindling detecting step further comprises:

Swindle parameter determining step, compares described swindle parameter and a threshold value, thereby judges whether described trading object exists the swindle of transaction rank.

39. methods according to claim 1, is characterized in that, described method also comprises:

Historical information obtaining step, the described historical information of at least one trading object described in obtaining.

40. according to the method described in claim 39, it is characterized in that, in described historical information obtaining step, from online store operator, obtains described historical information, or extracts described historical information from the data of online store issue.

41. methods according to claim 1, it is characterized in that, described historical information comprises historical ranking information, it is expressed as a rank sequence corresponding with discrete-time series, each element in described rank sequence, corresponding to a discrete time point in described time series, represents the rank of described trading object when described discrete time point.

42. methods according to claim 1, is characterized in that, described historical information comprises history evaluation information, the evaluation information that described in historical each time period, the user of trading object makes this trading object.

43. methods according to claim 1, is characterized in that, described historical information comprises historical review information, the user comment that described in historical each time period, the user of trading object makes described trading object.

44. methods according to claim 1, is characterized in that, described method also comprises: the described active period of detected described trading object is sent to online store operator or at least one network user.

45. methods according to claim 1, is characterized in that, described method also comprises: the result that transaction swindling detecting step is detected sends to online store operator or at least one network user.

In 46. 1 kinds of network tradings, bad user's detection system, is characterized in that, described system comprises:

Active period detecting unit, for detecting the active period of at least one trading object based on historical information;

Transaction swindling detecting unit, for based at least one evidence, the active period of described at least one trading object being detected, determines the trading object that has the swindle of transaction rank; And

Bad user's detecting unit, carries out at least one in the active period for detection of the trading object of swindling in described existence transaction rank and hands over easy-operating bad user.

47. according to the system described in claim 46, it is characterized in that, described active period detecting unit further comprises:

Enliven event checking module, for detect the event of enlivening of described at least one trading object based on described historical information;

Active period determination module, for merging the adjoining described event of enlivening to form the described active period of described at least one trading object.

48. according to the system described in claim 47, it is characterized in that, the described event of enlivening is that described trading object continues the higher time period of rank in trading object ranking list, and the standard that rank is higher is that the rank of described trading object in trading object ranking list is not more than a rank threshold k *.

49. according to the system described in claim 48, it is characterized in that, described system also comprises a rank threshold value setting unit, for described rank threshold k * is set.

50. according to the system described in claim 47, it is characterized in that, adjacent two time intervals of enlivening event are less than to an interval threshold φ as enlivening the standard of event merge in same active period by described two.

51. according to the system described in claim 50, it is characterized in that, described system also comprises an interval threshold setting unit, for described interval threshold φ is set.

52. according to the system described in claim 47, it is characterized in that, described, enlivens in event detection unit, further comprises:

Start time identification module, for identifying the described start time of enlivening event;

End time identification module, for identifying the described end time of enlivening the time;

Enliven event recognition module, for the time period between each start time and its afterwards adjacent end time is identified as to the event of enlivening, thereby detect all events of enlivening of described trading object.

53. according to the system described in claim 52, it is characterized in that,

Described start time identification module, for the rank of each the above trading object of time point of historical ranking information of historical information described in sequential search, when the rank that is not more than a rank threshold k * and a upper time point when the rank of current point in time is greater than described rank threshold k *, identification current point in time is the described described start time of enlivening event;

Described end time identification module, for the rank of each the above trading object of time point of historical ranking information described in sequential search, when the rank that is greater than described rank threshold k * and a upper time point when the rank of current point in time is not more than described rank threshold k *, identifying a upper time point is the described described end time of enlivening event.

54. according to the system described in claim 52, it is characterized in that,

Described start time identification module, for when the rank of first the above trading object of time point of the historical ranking information of described historical information is not more than described rank threshold k *, described first time point of identification is time at the beginning;

Described end time identification module, for when the rank of last the above trading object of time point of described historical ranking information is not more than described rank threshold k *, described last time point of identification is an end time.

55. according to the system described in claim 47, it is characterized in that,

Described active period detecting unit, initial time point for the historical ranking information from described historical information starts the each event of enlivening detecting of sequential search, when current, enliven event and upper one time interval of enlivening event while being less than an interval threshold φ, enliven event merge in same active period by these two, until searched for all events of enlivening that detect to detect all active period of described trading object.

56. according to the system described in claim 55, it is characterized in that,

Described active period detecting unit, for enlivening event and upper one time interval of enlivening event was not less than described interval threshold φ when one, and this enlivens event and next and enlivens the time interval of event while being not less than described interval threshold φ, detect this and enliven event from as an active period.

57. according to the system described in claim 46, it is characterized in that, described transaction swindling detecting step further comprises:

Evidence authentication module, for based at least one evidence, the active period of described at least one trading object being detected, and obtain each trading object one swindle parameter.

58. according to the system described in claim 57, it is characterized in that, described transaction swindling detecting unit further comprises:

Enliven event analysis module, for identifying at least one ascent stage, maintenance stage and decline stage of enlivening event of described active period.

59. according to the system described in claim 58, it is characterized in that, the described event analysis module of enlivening, for determining the described rank of enlivening trading object described in event first time and last time in a peak value range delta R, time period between described first time and described last time is identified as to the described maintenance stage, by described, enliven the time period before the maintenance stage in event and be identified as described ascent stage, by described, enliven the time period after the maintenance stage in event and be identified as the described decline stage.

60. according to the system described in claim 58, it is characterized in that, ascent stage and/or the decline stage of enlivening in event of described evidence based in described active period forms.

61. according to the system described in claim 58, it is characterized in that,

The maintenance stage of enlivening in event of described evidence based in described active period forms.

62. according to the system described in claim 57, it is characterized in that,

The quantity of described evidence based on enlivening event in described active period forms.

63. according to the system described in claim 57, it is characterized in that, average ratings and the historical average ratings of described evidence based in described active period forms.

64. according to the system described in claim 57, it is characterized in that,

In the distribution of described evidence opinion rating in described active period based on trading object and history evaluation information, the distribution of opinion rating forms.

65. according to the system described in claim 57, it is characterized in that, described evidence forms based on the similarity degree between user comment in described active period.

66. according to the system described in claim 57, it is characterized in that,

Described evidence based on the theme of trading object user comment in described active period, distribute and historical review information in the theme of the user comment formation that distributes.

67. according to the system described in claim 57, it is characterized in that, described evidence authentication module, for considering described at least one evidence, to based on described at least one evidence, verify that the correspondence swindle parameter obtaining is weighted, thereby obtain described swindle parameter.

68. according to the system described in any one in claim 57-67, it is characterized in that, described rank fraud detection unit further comprises:

Swindle parameter judge module, for described swindle parameter and a threshold value are compared, thereby judges whether described trading object exists the swindle of transaction rank.

69. according to the system described in claim 46, it is characterized in that, described system also comprises:

Historical information acquisition module, for obtaining the described historical information of described at least one trading object.

70. according to the system described in claim 69, it is characterized in that, described historical information acquiring unit for obtaining described historical information from online store operator, or extracts described historical information from the data of online store issue.

71. according to the system described in claim 46, it is characterized in that, described system also comprises an active period transmitting element, for the described active period of detected described trading object is sent to online store operator or at least one network user.

72. according to the system described in claim 46, it is characterized in that, described system also comprises a transaction rank fraud detection result transmitting element: result detected transaction swindling detecting unit is sent to online store operator or at least one network user.

In 73. 1 kinds of network tradings, bad user's detection method, is characterized in that, described method comprises:

Transaction swindling detecting step, detects the active period of at least one trading object based at least one evidence, determines the trading object that has the swindle of transaction rank;

Bad user's detecting step, detection is carried out at least one and is handed over easy-operating bad user in the active period of the trading object of described existence transaction rank swindle.

In 74. 1 kinds of network tradings, bad user's detection system, is characterized in that, described system comprises:

Transaction swindling detecting unit, for based at least one evidence, the active period of at least one trading object being detected, determines the trading object that has the swindle of transaction rank;

Bad user's detecting unit, carries out at least one in the active period for detection of the trading object of swindling in described existence transaction rank and hands over easy-operating bad user.

Images