JP2008015989A - Network transaction wrongdoer detecting method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a network transaction wrongdoer detecting method for detecting fraudulent evaluation in advance to effectively prevent a fraudulent act performed on the extension. <P>SOLUTION: A method of detecting a wrongdoer in a network system providing a function of electronically performing transaction and a function of performing evaluation by a transactor after transaction, and enabling a user to browse an evaluated result, comprises a divergence value computing process for computing divergence values between actual evaluation point accumulated total values within a latest fixed period and values when the evaluation point accumulated total values change on the average in the meantime, for every evaluated person, a fraudulent evaluation determination process for comparing the divergence values with a predetermined value and determining that there is fraudulent evaluation when the divergence values are the predetermined value or more, and a fraudulent evaluated person extracting process for extracting the evaluated person determined to have fraudulent evaluation, as a candidate of a person evaluated as fraudulent. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a network transaction fraud detection method in a network auction, a network shopping, a personal transaction, etc. in which transactions are performed via a network such as the Internet.

  Recently, the Internet auction system on the Internet has been actively used, but because it is an electronic transaction, the transaction is conducted in a state where the other party's face is not visible and the actual product cannot be confirmed directly. Therefore, in order to increase the safety of transactions, a mutual evaluation system has been introduced, and it is possible to determine whether or not to conduct a transaction by referring to the past evaluation of the transaction partner during the transaction (for example, , See Non-Patent Document 1.)

  That is, after finishing the transaction, the seller evaluates the winning bidder who is the trading partner, and the winning bidder evaluates the seller who is the trading partner. FIG. 1 is a diagram illustrating an example of evaluation in an online auction, in which five-level evaluations of “very good”, “good”, “neither”, “bad”, and “very bad” are performed. In addition, each evaluation has “+1” for “very good” and “good”, “0” for “neither”, “−1” for “bad” and “very bad”. The accumulated evaluation point value is managed for each user (exhibitor, successful bidder) ID, and can be referred to at the time of transaction. The accumulated evaluation point value is counted according to the number of people evaluated, not the number of evaluations. Even if multiple evaluations are made in one transaction, only the latest evaluation is reflected, and the same person Only the latest evaluation reflects the evaluation even if you make multiple transactions. In addition to such quantitative evaluation, qualitative evaluation based on comments such as impressions on transactions can be performed.

  FIG. 2 is a diagram showing a display example of evaluation, and FIG. 2A shows a “very good / good” evaluation point cumulative value, a “very bad / bad” evaluation point cumulative value for a predetermined user, A total evaluation point cumulative value obtained by subtracting a “very bad / bad” evaluation point cumulative value from a “very good / good” evaluation point cumulative value is displayed. (B) displays an exhibitor ID and a total evaluation point accumulated value. (C) displays the evaluation by successful bid and the evaluation by exhibition for a predetermined user. (D) displays the accumulated evaluation point value for each period for a predetermined user.

Such an evaluation system can be used not only for online auctions but also for online shopping and personal transactions.
H. Takahasi, Y. Usui and N. Yoshikai: Evaluation of Reputation System in Internet Auctions -Experimental Approach, NTT Technical Review, Vol.2, No.6 (2004).

  As described above, in online auctions, etc., user evaluation is practically important for ensuring the safety of transactions, but fraudulent acts called “illegal evaluations” that exploit this point are problematic. Yes.

  FIG. 3 is a diagram showing an outline of fraud evaluation, and accumulating evaluation points by evaluating “very good / good” from a plurality of fraud evaluation IDs forming a fraud evaluation community to a plurality of fraud evaluation IDs. A fraudulent act (such as acquiring only the price and not delivering the product) is performed against a general user who has raised the value unfairly and believed in the evaluation. Although the same person may have a large number of fraud evaluation IDs and fraudulent evaluation IDs, it may cost a little to acquire IDs. Therefore, fraud evaluation is performed with a small expense by forming a fraud evaluation community. There are many cases.

  In the Internet auction system, there has been a problem of “fictitious name bidding” (an act of bidding on a product that I submitted using another fictitious ID and lifting the winning bid amount). A lot of research has been done, but not enough research has been done on the above-mentioned “fraud assessment”.

  The present invention has been proposed in view of the above-mentioned conventional problems, and the object of the present invention is to detect a fraud evaluation in advance and effectively prevent fraudulent acts performed on the extension. The object is to provide a method for detecting fraudulent transactions.

  In order to solve the above-mentioned problems, in the present invention, as described in claim 1, a function for electronically conducting a transaction and a function for a trader to evaluate after the transaction are provided, and the evaluation result Is a method for detecting fraudsters in a network system that can be viewed by users, and the cumulative total of evaluation points during the most recent fixed period, and the average evaluation point has changed during that period. Deviation value calculation step for calculating a deviation value from the case value for each person to be evaluated, and fraud evaluation for judging that fraud evaluation has occurred when the deviation value is equal to or greater than a predetermined value by comparing the deviation value with a predetermined value. The gist of the present invention is a network transaction fraudster detection method comprising a judgment step and a fraudulent evaluator extraction step of extracting a to-be-evaluated person who has been judged to have been fraudulent as a fraudulent evaluator candidate.

  In addition, as described in claim 2, in the network transaction fraudster detection method according to claim 1, the network system includes a function of electronically performing exhibition, bidding and successful bid, It may be a net auction system that provides a function for successful bidders to perform mutual evaluation.

  Moreover, as described in claim 3, in the network transaction fraudster detection method according to claim 1, the divergence value calculating step and the fraud evaluation determining step divide the window size w which is a processing target period. It can be executed every period divided by several L.

  Moreover, as described in claim 4, in the network transaction fraudster detection method according to claim 3, the divergence value calculation step includes an actual value on each processing completion date of the divergence value calculation within the processing target period. A vector with the evaluation point accumulated value as an element, and a vector with the element when the evaluation point accumulated value changes on the average on each processing completion date of the deviation value calculation within the processing target period Thus, the deviation value can be calculated by integrating the absolute values of the differences of the corresponding elements of both vectors.

  In addition, as described in claim 5, in the network transaction fraudster detection method according to claim 2, the store registration that has been examined in advance by the operator of the net auction system as the processing target of the divergence value calculation step. Can be excluded.

  Further, as described in claim 6, in the network transaction fraudster detection method according to claim 1, a group of fraud assessors from the group of fraud assessors extracted by the fraud assessor extraction step. A fraud evaluator group extraction step for extracting the fraud evaluation, and a visualization step for visualizing the correlation between the fraud evaluations from the extracted group of fraudulent evaluators and the group of fraud evaluators.

  In addition, as described in claim 7, fraudulent acts in a network system that provides a function for conducting a transaction electronically and a function for a trader to perform an evaluation after the transaction, and enabling the user to view the evaluation result This is a device for detecting a person, and for each person to be evaluated, a divergence value between an actual evaluation point cumulative value and an average evaluation point change value during the most recent fixed period. A divergence value calculating means for calculating the divergence value, a fraud evaluation judging means for judging that there is a fraud evaluation when the divergence value is equal to or greater than a predetermined value by comparing with the predetermined value, and a fraud evaluation is determined It can be configured as a fraudster detection device comprising fraudulent evaluator extraction means for extracting the evaluated evaluator as a fraudulent evaluator candidate.

  In the network transaction fraudulent person detection method of the present invention, fraud evaluation can be detected in advance, and fraudulent acts performed on the extension can be effectively prevented.

  Hereinafter, preferred embodiments of the present invention will be described. In addition, although an internet auction will be described as an example, the present invention can be applied to other general transactions using networks such as online shopping and personal transactions.

<Principle of fraud assessment detection>
FIG. 4 is a diagram illustrating an example of changes in the accumulated evaluation point value of an exhibitor with and without fraud evaluation. In FIG. 4, when there is no fraud evaluation, if the accumulated evaluation point value increases with the passage of time as shown by a broken line, if there is fraud evaluation, a short period (one month or Less than that), the cumulative value of evaluation points suddenly increases, and if it increases to a certain extent, the rapid increase stops. Since this is a preparatory stage for conducting commercial fraud, it is intended to increase the cumulative total of evaluation points in a short period of time, and there will be no further increase when the total cumulative evaluation point is reached. This is due to meaning or what is seen as unnatural.

  In the present invention, paying attention to such a property of the evaluation point accumulated value, as shown in FIG. 5, the actual evaluation point accumulated value p in the most recent fixed period and the evaluation point accumulated value averagely change during the period. In this case, a deviation value from the value d is calculated, and when the deviation value is equal to or greater than a predetermined value, it is determined that there has been a fraud evaluation. In addition, there is a method to judge fraud when the rate of change of the accumulated evaluation point value exceeds a predetermined value. However, as shown in FIG. 6, the rate of change may not be so large as shown in FIG. From the viewpoint of preventing detection omission in such a case, the method based on the divergence value is excellent.

  In addition, when implementing the above-mentioned principle, a large amount of actual data must be processed in real time with limited computer resources, so calculation of divergence values with a large amount of calculation and comparison with predetermined values are performed at predetermined intervals. (It will be described later).

<System configuration>
FIG. 7 is a diagram showing a configuration example of a system to which the present invention is applied.

  In FIG. 7, a network 2 such as the Internet is connected to a plurality of terminals 1 such as PCs (Personal Computers) and mobile phones, and the network 2 is connected to a net auction system 3 for managing and operating a net auction. The system 3 is connected to a fraudster detection device 4 that detects a user who is likely to perform fraud evaluation.

  The net auction system 3 includes an exhibition / bid / successful bid processing unit 31 that performs auction / bidding / successful bid processing for a net auction, and a mutual evaluation processing unit 32 that performs an evaluation process mutually performed by the exhibitor and the successful bidder after the transaction. And an auction DB (Data Base) 33 for accumulating data relating to transactions and data related to evaluation (evaluation log data).

  The fraudster detection device 4 obtains evaluation log data from the auction DB 33 of the net auction system 3 and analyzes it to extract a fraudulent evaluation ID (user ID that is fraudulently evaluated). The fraud evaluation community extraction unit 42 extracts the fraud evaluation ID group and the fraud evaluation ID group using the fraud evaluation ID extracted by the fraud evaluation ID extraction unit 41 as a seed, and extracts the fraud evaluation community. And a data holding unit 43 that holds data and output data required in the course of processing.

<Configuration and operation of unauthorized evaluation ID extraction unit>
FIG. 8 is a diagram illustrating a configuration example of the unauthorized evaluation ID extraction unit 41.

  In FIG. 8, the fraudulent evaluation ID extraction unit 41 includes a control unit 411 that performs overall control in the fraudulent evaluation ID extraction unit 41 and a data stream newly generated in the auction DB 33 of the net auction system 3. The evaluation log data selection unit 412 that selects the evaluation log data to be analyzed, and the evaluation log data of the seller who has received store registration by the preliminary examination of the operator of the net auction system 3 is excluded from the evaluation log data. The store registration ID exclusion unit 413, the seller evaluation selection unit 414 that excludes the evaluation for the successful bidder and selects only the evaluation log data for the seller, and excludes “very bad / bad” and “neither” Positive evaluation selection unit 415 that selects only “very good / good” evaluation log data and excluded / selected And a evaluation after the evaluation point total value updating section 416 for updating to the latest state evaluation point cumulative value based on the evaluation log data. It should be noted that the reason why the evaluation log data relating to the seller who has received the store registration is excluded is that it has been examined in advance and it is unlikely that it will be illegal. The reason why only the evaluation log data for the seller is selected is that the evaluation for the seller is necessary for the analysis of the fraud evaluation, and the evaluation for the successful bidder is unnecessary. The reason why only the evaluation log data that is “very good / good” is selected is that it is not considered that the evaluation is “very bad / bad” or “neither” by fraud.

  Further, the fraudulent evaluation ID extraction unit 41 periodically calculates an evaluation point cumulative value vector and a deviation value, which will be described later, and an evaluation point cumulative value vector / deviation value update unit 417, and the deviation value exceeds a predetermined value. A divergence value determination unit 418 that determines whether or not the divergence value exceeds a predetermined value, and a fraud evaluation ID extraction unit 419 that outputs a user ID corresponding to the divergence value as a fraud evaluation ID.

  FIG. 9 is a diagram illustrating an example of evaluation log data acquired from the auction DB 33 of the net auction system 3. “Evaluation date / time” “Evaluation target (exhibitor / successful bidder)” “Evaluator ID” “Evaluated ID” It includes items such as “evaluation points”, “evaluation points accumulated before evaluation”, and “evaluation points accumulated after evaluation”.

  FIG. 10 is a diagram showing an example of data held for each ID in the data holding unit 43, an actual evaluation point cumulative value vector, a vector of evaluation point cumulative values assuming that the average has increased, It includes the deviation value between the actual evaluation point cumulative value and the evaluation point cumulative value assuming that the average increase has occurred, and the current evaluation point cumulative value. The meaning of each data will be described below.

  In FIG. 11, N is the current date, T is the processing completion date such as the latest divergence value calculation, w is the window size that is the processing target period, T-w is the processing start date, and only the window size w from T is shown. The retroactive period (Tw to T) is the period that is the target of processing in T. The processing completion date immediately before T is T-w / L before w / L, where L is the number of divisions, and follows the w / L interval.

  Here, the actual evaluation point accumulated value vector is defined as a vector having elements of the actual evaluation point accumulated value p on each processing completion date within the processing target period. This can be expressed as follows.

平均増加したと仮定した場合の評価ポイント累計値のベクトルは、処理対象期間内の各処理完了日における、平均的に評価ポイント累計値が変化したとした場合の値ｄを要素にもったベクトルとして定義される。式で示せば次のようになる。

Assuming that the average increase has occurred, the vector of accumulated evaluation points is a vector having elements as values d when the average accumulated evaluation points change on each processing completion date within the processing target period. Defined. This can be expressed as follows.

なお、任意の日ｔにおける値ｄは次式で計算できる。

Note that the value d on an arbitrary date t can be calculated by the following equation.

また、実際の評価ポイント累計値と平均増加したと仮定した場合の評価ポイント累計値の乖離値は、実際の評価ポイント累計値のベクトルと平均増加したと仮定した場合の評価ポイント累計値のベクトルの対応する要素毎の差の絶対値を積算したものとして定義される。式で示せば次のようになる。

In addition, the difference between the actual evaluation point cumulative value and the average evaluation point cumulative value assuming that the average increase has occurred is the difference between the actual evaluation point cumulative value vector and the evaluation point cumulative value vector assuming that the average increase has occurred. It is defined as the sum of the absolute values of the differences for the corresponding elements. This can be expressed as follows.

図１２は被不正評価ＩＤ抽出部４１の処理例を示すフローチャートである。

FIG. 12 is a flowchart illustrating a processing example of the unauthorized evaluation ID extraction unit 41.

  In FIG. 12, assuming that the processing is periodically started at an appropriate timing every day and the current date is N (step S101), the difference between the current date N and the latest processing completion date T is w / L. It is determined whether or not processing such as divergence value calculation should be executed depending on whether or not it is smaller (step S102).

  When the difference between the current date N and the latest fraud determination processing completion date T is smaller than w / L (Yes in step S102), the stream data acquired from the auction DB 33 of the net auction system 3 is repeatedly processed (step S103). In this iterative process, the evaluation log data (FIG. 9) is selected from the data stream (step S104), and whether or not the “evaluated person ID” is an ID not registered in the store (step S105). ), Whether or not the “evaluation target” is an exhibitor (step S106), and whether or not the “evaluation point” is a value (“+1”) corresponding to “very good / good” (step S107) To do. When the ID is registered in the store (No at Step S105), when the evaluation target is not an exhibitor (No at Step S106), when the evaluation point does not correspond to “very good / good” (No at Step S107) Deletes the acquired evaluation log data (step S108) and returns to the iterative process (step S103).

  Next, it is determined whether or not the evaluation log data that has not been deleted is a target that exists in the ID that already holds the data in the data holding unit 43 (step S109), and if it exists (Yes in step S109). ) Calls the corresponding data (FIG. 10) (step S110), and if it does not exist (No in step S109), creates new data (step S111).

Next, evaluate the "evaluation point cumulative value after the evaluation" of log data, and assigned to the current evaluation point total value P _N of the calling data or data newly created (step S112).

  Then, when the repetitive processing is completed for all data in the data stream, the processing ends (steps S113 and S119).

On the other hand, when the difference between the current date N and the latest processing completion date T is equal to w / L at the start of processing (No in step S102), _PN in the data of each ID held in the data holding unit 43 is D _N (step S114), defining a vector of actual evaluation point accumulated values and a vector of evaluation point accumulated values when assuming an average increase, and assuming an average increase with the actual evaluation point accumulated value The divergence value of the accumulated evaluation point values is calculated (step S115). The range of each vector and divergence value is from T- (L-2) w / L to N since the oldest vector element is deleted.

  Next, it is determined whether or not the deviation value of the evaluation point accumulated value when the average evaluation point accumulated value and the average increase are smaller than the predetermined value Q (step S116). T of the calculated data is set to N and held in the data holding unit 43 (step S117), and the process ends (step S119).

  Further, when the deviation value of the evaluation point accumulated value when the average evaluation point accumulated value and the average increase are equal to or larger than the predetermined value Q (No in step S116), the ID is extracted as an unauthorized evaluation ID ( The data is output to a predetermined area in the data holding unit 43 (step S118), and the process is terminated (step S119).

<Configuration and operation of the fraud evaluation community extractor>
FIG. 13 is a diagram illustrating a configuration example of the fraud evaluation community extraction unit 42.

  In FIG. 13, the fraud evaluation community extraction unit 42 includes a control unit 421 that performs overall control in the fraud evaluation community extraction unit 42, a set initialization unit 422 that initializes a set (data) used for processing, and a candidate A candidate fraud evaluation ID set update unit 423 that updates the fraud evaluation ID set, a fraud evaluation ID set update unit 424 that updates the fraud evaluation ID set, and a fraud evaluation ID set update unit 425 that updates the fraud evaluation ID set; The fraudulent evaluation ID group / fraud evaluation ID group output unit 426 that outputs the fraudulent evaluation ID group / fraud evaluation ID group based on the data included in each set, and fraud from the fraud evaluation ID group / fraud evaluation ID group A fraud evaluation community visualization unit 427 that can visually display a fraud evaluation community that performs evaluation.

  FIG. 14 is a flowchart illustrating a processing example of the fraud evaluation community extraction unit 42.

In FIG. 14, when the process is started (step S201), the candidate fraud evaluation ID set F _candidate is initialized to an empty set (F _candidate = {φ}) (step S202).

  Next, the unauthorized evaluation ID set C is initialized by substituting the element s of the unauthorized evaluation ID extracted by the unauthorized evaluation ID extraction unit 41 as a seed (C = {s}) (step S203).

  Next, the fraud evaluation ID set F is initialized to an empty set (F = {φ}) (step S204).

Next, OR, which is all IDs evaluating the element s of the unauthorized evaluation ID set C, is OR-joined with the candidate unauthorized evaluation ID set F _candidate (F _candidate = F _candidate ∪ {u}) (step S205).

Next, all the v evaluated IDs evaluated by p or more IDs among the IDs belonging to the candidate fraud evaluation ID set F _candidate are extracted and ORed with the fraud evaluation ID set C (C = C ∪ {v}) (step S206).

Next, the ID set in F _candidate that evaluates an arbitrary ID included in the fraud evaluation ID set C is substituted into the fraud evaluation ID set F (step S207).

  Next, the unauthorized evaluation ID set C is output as an unauthorized evaluation ID group of the unauthorized evaluation community, and the unauthorized evaluation ID set F is output as an unauthorized evaluation ID group (step S208).

  Then, the fraud evaluation community is visualized from the reference relationship between the IDs of the fraudulent evaluation ID group and the fraud evaluation ID group (step S209), and the process ends (step S210).

  For example, as shown in FIG. 15, a plurality of object graphics corresponding to IDs belonging to the fraud evaluation ID group are arranged in the center of the screen, and a plurality of object graphics corresponding to IDs belonging to the fraud evaluation ID group are arranged around it. Then, the evaluation (illegal evaluation) for the ID belonging to the unauthorized evaluation ID group from the ID belonging to the unauthorized evaluation ID group can be visualized by expressing a line with an arrow between the corresponding object figures. .

<Summary>
As described above, the present invention pays attention to the fact that when there is a fraud evaluation, the accumulated evaluation point value rapidly increases in a short period of time, and when it increases to a certain extent, the rapid increase stops, In this case, the deviation value between the actual accumulated evaluation point value and the value when the accumulated evaluation point value is averagely changed in the meantime is obtained. Since it is determined that there is a user ID, it is possible to detect with high accuracy a user ID that is highly likely to be fraudulently evaluated. The window size w = 30 days for the actual operation log data set (approximately 3.6 million) from April 7, 2005 to September 7, 2005 at "Yahoo! Auction" operated by Yahoo Japan Corporation. When the method of the present invention was applied with the number of window divisions L = 6, it was confirmed that violation history exists in 549 out of the top 1000 extracted fraud evaluation IDs, confirming the usefulness of the present invention. did it.

  In addition, the identified IDs to be illegally evaluated or individual IDs of the ID groups for illegal evaluation are sorted and listed in descending order of possibility of illegal evaluation (in descending order of divergence value, etc.) for artificial confirmation. Through the above, fraudulent acts following fraud evaluation can be effectively prevented by making it a target of fraud prevention measures (suspension of use, etc.).

  The present invention has been described above by the preferred embodiments of the present invention. While the invention has been described with reference to specific embodiments, various modifications and changes may be made to the embodiments without departing from the broad spirit and scope of the invention as defined in the claims. Obviously you can. In other words, the present invention should not be construed as being limited by the details of the specific examples and the accompanying drawings.

It is a figure which shows the example of evaluation in a net auction. It is a figure which shows the example of a display of evaluation. It is a figure which shows the outline | summary of fraud evaluation. It is a figure which shows the example of the change of an evaluation point total value of an exhibitor with and without a fraud evaluation. It is explanatory drawing of the deviation value of the change of an actual evaluation point cumulative value, and an average change. It is a figure which shows an example in case a deviation value is large, but a rate of change does not become large. It is a figure which shows the structural example of the system to which this invention is applied. It is a figure which shows the structural example of an unauthorized evaluation ID extraction part. It is a figure which shows the example of evaluation log data. It is a figure which shows the example of the data hold | maintained with respect to each ID. It is explanatory drawing of an evaluation point accumulated value vector and a deviation value. It is a flowchart which shows the process example of a to-be-unauthorized evaluation ID extraction part. It is a figure which shows the structural example of a fraud evaluation community extraction part. It is a flowchart which shows the process example of a fraud evaluation community extraction part. It is a figure which shows the example of the fraud evaluation community visualized.

Explanation of symbols

1 terminal 2 network 3 net auction system 31 exhibition / bid / successful bid processing unit 32 mutual evaluation processing unit 33 auction DB
DESCRIPTION OF SYMBOLS 4 Cheating person detection apparatus 41 Fraudulent evaluation ID extraction part 411 Control part 412 Evaluation log data selection part 413 Store registration ID exclusion part 414 Exhibitor evaluation selection part 415 Positive evaluation selection part 416 Evaluation point accumulated value update part 417 after evaluation Evaluation Point Cumulative Value Vector / Deviation Value Update Unit 418 Deviation Value Judgment Unit 419 Unauthorized Evaluation ID Extraction Unit 42 Unauthorized Evaluation Community Extraction Unit 421 Control Unit 422 Set Initialization Unit 423 Candidate Unauthorized Evaluation ID Set Update Unit 424 Unauthorized Evaluation ID Set Update unit 425 Unauthorized evaluation ID set update unit 426 Unauthorized evaluation ID group / Unauthorized evaluation ID group output unit 427 Unauthorized evaluation community visualization unit 43 Data holding unit

Claims (7)

A method for detecting fraudsters in a network system that provides a function for conducting a transaction electronically and a function for a trader to perform an evaluation after the transaction, and enables the user to view the evaluation result,
A divergence value calculation step for calculating a divergence value between the actual evaluation point accumulated value in the most recent fixed period and a value when the evaluation point accumulated value changes on average during that period,
A fraud evaluation judging step of judging that there has been a fraud evaluation when the divergence value is compared with a predetermined value and becomes a predetermined value or more;
A network transaction fraudster detection method comprising: a fraudulent evaluator extraction step of extracting an evaluator determined to have been fraudulent evaluation as a fraudulent evaluator candidate. In the network transaction fraudster detection method according to claim 1,
The above network system is a network auction system that provides a function of electronically performing exhibition, bidding and successful bid, and a function of mutual evaluation between the seller and the successful bidder after the transaction. Person detection method. In the network transaction fraudster detection method according to claim 1,
The network transaction fraudster detection method, wherein the divergence value calculation step and the fraud evaluation determination step are executed every period obtained by dividing the window size w that is a processing target period by the division number L. The network transaction fraudster detection method according to claim 3,
The above deviation value calculation step
A vector with elements of the actual accumulated value of evaluation points on each processing completion date for calculating the divergence value within the processing target period,
From the vector that has the value when the evaluation point cumulative value changes on the average on each processing completion date of calculation of the deviation value within the processing target period,
A network transaction fraudster detection method, characterized in that a divergence value is calculated by integrating absolute values of differences between corresponding elements of both vectors. The network transaction fraudster detection method according to claim 2,
A network transaction fraudster detection method characterized by excluding store registrants previously screened by an operator of a net auction system as a processing target of the divergence value calculation step. In the network transaction fraudster detection method according to claim 1,
A fraud evaluator group extraction step for extracting a group of fraud evaluators from the group of fraud rate evaluators extracted by the fraud evaluator extraction step;
A network transaction fraudster detection method comprising: an extracted fraudster group and a visualization step of visualizing a correlation between fraud evaluations from the fraudster group. A device that detects a fraudulent person in a network system that provides a function for conducting a transaction electronically and a function for a trader to perform an evaluation after the transaction, and enables the user to view the evaluation result,
Deviation value calculation means for calculating, for each person to be evaluated, a deviation value between the actual accumulated evaluation point value in the most recent fixed period and the value when the accumulated evaluation point value has changed in the meantime,
A fraud evaluation judging means for judging that there has been a fraud evaluation when the divergence value is equal to or greater than a predetermined value by comparing with the predetermined value;
A fraudster detection apparatus comprising: a fraudulent evaluator extracting unit that extracts a to-be-evaluated person who has been judged to have been fraudulently evaluated as a fraudulent evaluator candidate.

Images