AU2007202119A1 - A fraud detection filter - Google Patents

Description

P/00/011 Regulalion 3.2

AUSTRALIA

Patents Act 1990 COMPLETE SPECIFICATION STANDARD PATENT

(ORIGINAL)

Name of Applicant: Actual Inventors: Address for Service: Invention Title: Fraud Management Technologies Pty Ltd, ABN 88 117 945 945, of Level 50, 101 Collins Street, Melbourne, Victoria, 3000 Constantine Siourthas Bjarne Staugaard Matzen DAVIES COLLISON CAVE, Patent Attorneys, of 1 Nicholson Street, Melbourne, Victoria 3000, Australia 'A fraud detection filter' The following statement is a full description of this invention, including the best method of performing it known to us:- P \OPER\ 8I)19I II.. Clo-uOp A L d I I A,52W I A FRAUD DETECTION FILTER

FIELD

The present invention relates to a system for detecting possible fraudulent activity during an access request process in an online environment, and in particular to a fraud detection filter.

BACKGROUND

Current systems for protection against fraud in online environments are deficient. Web server systems currently employ authentication systems to authenticate users when they request access to connect to and use server applications. The authentication systems seek to determine that an access request is being made by an authorised user, e.g. on the basis of a unique username and password combination or some other unique authentication data submitted biometric data). However no attempt is made to determine if the data submitted indicates the authentication data has been compromised or another party is fraudulently using the authorised parties' data or access privileges before the connection to the server application is allowed.

Online banking systems, for example, authenticate users, establish a connection session and allow transactions with an Internet banking application to be completed during the session; fraud detection is only performed subsequently by back-end analytic systems. The analytic systems receive transaction data of the session and process the data for comparison with pattern data representing possible fraudulent conditions. This is clearly unsatisfactory as a user's account can be compromised before any fraud is detected.

Suspicious activities or other undesirable conditions may not be detected until identified by the back-end analytic software, i.e. after the event has occurred.

It is desired to address the above, or at least provide a useful alternative.

P \OPERDfRIW IT IH I ]Plopu$Cmnp AUdoc- ;2(~f7 -2-

SUMMARY

In accordance with the present invention there is provided a fraud detection filter installed in an application server including a secure application, said filter including a rules engine for receiving request data representing an access request for the secure application from a N? user, and applying at least one risk condition rule to the request data for generating a risk probability level, and detecting at least one fraud condition when said risk probability level exceeds a threshold level, before passing said access request to said secure application.

Advantageously the rules engine may access past session data for said user in applying said at least one risk condition rule. Also, the rules engine may advantageously access application data accessed by said application for said user in applying said at least one risk condition rule.

Preferably applying the fraud condition rule includes determining a location associated with an Internet Protocol address associated with the access request. Preferably applying the fraud condition rule includes determining a change in the location of the Internet Protocol address of a user associated with the access request since receiving a previous access request from the user.

Preferably, the process includes generating an action command based on the risk probability level and risk policy data. Preferably the risk policy data represents said threshold level. The action command may invoke a two-factor authentication process for confirming the identity of the user or a denial of access for the user.

The present invention also provides a management server for generating an interface to adjust and set said at least one risk condition rule.

The present invention also provides a filter system including the filter and the management server.

P OPERlDRVA1 (J I I S MIwSCuoop AU doc I IA)5/2(9)7 -3- The present invention also provides an application server including a secure application for access by a user and the fraud detection filter.

The present invention also provides a process for detecting a fraud condition, performed by an application server, including: receiving request data representing an access request from a user for a secure application of the application server; applying at least one risk condition rule to the request data for generating a risk probability level; and detecting said fraud condition when said risk probability level exceeds a threshold level, before granting access to said secure application.

DESCRIPTION OF THE DRAWINGS Preferred embodiments of the invention are further described, by way of example only, with reference to the accompanying drawings, wherein: Figure 1 is a schematic diagram of a filter system arranged in a condition of use; Figure 2 is a schematic diagram of the filter system showing modules of an application server; Figure 3 is a flow chart of a filter process of the filter system Figure 4 is a flow chart of part of the filter process showing process blocks; Figure 5 is a schematic diagram of the filter system showing modules of a preprocessing filter and of a management server; Figure 6 is an image of a user interface showing user options; and Figure 7 is an image of a user interface generated by the management server of the filter system showing example process blocks.

DETAILED DESCRIPTION A fraud detection system in the form of a filter system 100, as shown in Figure 1, filters an access request from a user device 102 of a user 104 who is attempting to access a secure P \OPER\DBV IA3 1)I IS X A,,opu p AU do. I IfAS/2-7 -4application 204 on an application server 106. The user device 102 accesses the application server 106 via a first data network 108 the Intemrnet) and an associated network server 110 a Web server). The access request is filtered in the filter system 100 by a preprocessing filter 202 that is installed or embedded in the application server 106, as shown in Figure 2, before access is granted to the secure application 204.

The pre-processing filter 202 provides a real-time decision engine which performs blocking and alerting process actions depending on a risk probability level determined from the access request. The access request is in accordance with standard communication protocols, such as the suite of IP protocols, and may be a HTTP Get request. The action taken by the pre-processing filter 202, and the treatment of data obtainable from the access request, is configurable via a management server 112. The management server 112 is controlled from a management console 114 securely in communication with the management server 112, which is operated by an administrator 116. The administrator 116 may be the owner of the application server 106 and secure application 204 a bank owning an on-line banking application), or a third party security administrator providing security services to the owner of the secure application 204. The administrator 116 may conveniently configure and adapt the pre-processing filter 202 using the management console 114. The pre-processing filter 202 has access to an internal database 118 of the application server 106, for securely storing relevant filter data, and to a second data network 120 an intranet or the Internet), by which one or more external databases 122, a client device 124 (of a person client) 126 authorised by the owner to monitor performance of the filter 202) and the user devices 102 may be accessed.

The internal database 118 is used by the pre-processing filter 202 to keep a secure record of filter data, such as a history of past access requests and other data that may be deemed relevant by the administrator 116. By locating the filter 202 with the secure application 204 in the application server, the filter 202 is able to rely upon and use the same data and procedure calls as the secure application 204. The filter 202 can therefore access account data and access history data for individual user accounts on a per user level.

P \OPER\DIBVAll' I I o Coup AU do I I l I 2(W The pre-processing filter 202 may access the external database 122 to draw on third party information (such as published Internet Protocol address blacklists), or to deliver report data into a case management file. Report data may also be sent by the pre-processing filter 202 and the management server 112 to the client device 124 for real-time reporting by the filter system 100: for example, a person 126 may be alerted when the access request originates from a certain undesirable range of IP addresses.

The pre-processing filter 202 may access the user device 102 via the web server 110 and the first data network 108 to seek a second authentication factor. For example, the filter system 100 may request an additional user/password from the user 104, or submission of an encoded token sent by SMS to a mobile telephone, when certain access request characteristics, i.e. a fraud condition, is detected by processing of the access request by the pre-processing filter 202. In other words, a two factor authentication process can be invoked that needs to be satisfactorily completed before access is granted.

A filter process 300 performed by the pre-processing filter 202 commences with the preprocessing filter 202 receiving an access request sent by the user device 102 requesting access to the secure application 204 for the user 104 (step 302 in Figure Following reception of the access request, the pre-processing filter 202 then applies rules to the data of the request to generate the risk probability level, i.e. a measure representative of the probability that the access request originates from a risky, or fraudulent, user 104. Once the risk probability level is generated in step 304, the pre-processing filter 202 generates an action command depending on the level of the risk probability. For example, if there is a very low risk probability (determined at step 304) the action command (generated in step 306) may allow access to the secure application 204. On the other hand, if the risk probability is above an acceptable level, the action command may block access to the secure application 204 and/or alert the client 126 using a message sent to the client device 124. The filter process may continue by repeating the application of the rule/s (step 304) and the generation of a consequent action command (step 306) depending on the number of steps in the pre-processing filter 202. The owner 116 configures how many rules are applied and how many action commands are generated. The filter process ends once no P \OPER\BW AoI I l KoOpus CoMp AU do-I I/ (2X)7 -6further rules remain to be applied. For example, an access request including access request data may be received (at step 302) from an IP address located in an untrustworthy Russian city. The pre-processing filter 202 then applies a rule that considers the risk probability level associated with IP addresses from certain geographic locations, and assigns a relatively high risk level to this access request at step 304. Consequently, in step 306, the pre-processing filter 202 generates an action command to block the access request from the secure application 204 and a second action command to retain a record of this access request in the internal database 118.

Typical access request data, as analysed in step 302 of the filter process 300) may include one or more of the following characteristics of interest which may indicate a risk probability associated with the access request: Originating IP address, Browser type and version Connection speed Access from a known compromised IP address Access from public hot-spot Access via satellite Secure application data submitted with the request (such as account number, amounts, address changes etc.) The rules applied in step 304 relate to the access request data. Typical rules may include: Checking the originating country for the IP address is not a high risk or black listed country Impossible travel speed between current originating IP address and previous originating IP address Checking the originating IP address against known black lists Checking for money transfers to suspicious names. The filter 202, being part of the application server 106, applies rules to every access request made during a transaction session, even once a user has been given access privileges to the P \OPER\DBW 3I(9I Il 1l8ops cnp AU dm-]I A)5/2(X -7secure application 104. Accordingly, the rules are tailored for the specific application as required.

Checking if browser characteristics have changed from a previous request In an example of the filter process 300, the access request data is received in step 402 (shown in Figure In this example, the access request is for a Web application, and the access request data includes: data representative of the version of the Web browser used on the user device 102; and data representative of the IP address used by the user device 102.

Furthermore, this user 104 (identified by a username and password combination in the access request data) has previously interacted with the pre-processing filter 202, and thus historical data of previous access requests for other sessions is stored in the internal database 118. The first rule applied by the pre-processing filter 202 is a browser change rule (in step 404): if the browser version of the present access request has not changed since the previous access request, or is a more recent version, no action is taken by the preprocessing filter 202, and the filter process 300 continues to apply a second rule, being a land speed rule (in step 406). If, on the other hand, a downgrade of the browser version is detected (in step 404), a non-zero risk probability level is generated, and the pre-processing filter 202 generates an action command depending on the level of the fraud probability. If the risk probability level is high, corresponding to receipt of a percentage (say greater than of transactions, i.e. access requests, for a period that represent a browser downgrade, then an email alert action command is generated which leads to an email alert notice to be sent once to the client device 124.

If the risk probability level generated by the browser change rule (in step 404) is medium or low, the pre-processing filter 202 continues with an annotation action command being generated in step 410. The annotation action command tags record data in the internal database 118 to indicate that the access request data is suspect or dangerous has a corresponding risk probability level). If no browser downgrade was detected in step 404, a land speed rule (step 406) and a value comparer rule (step 408) are used by the preprocessing filter 202 to determine whether the present IP address is at a time and location which is greater than 400 km/h from the previous IP address and location that a user P \OPER DI W I 191 I IX lops oip AU d -I I IflS,I)7 -8- 104 would have had to have travelled at least 400 km/h to move between the previous IP address location and the current IP address location). If the land speed is less than 400 km/h, a low fraud probability is generated, and the pre-processing filter 202 generates an action command indicating that the access request data is and thus grants access to the secure application 204 (step 412). If, on the other hand, the user 104 appears to have travelled faster than 400 km/h, an action command is issued (at step 410) to annotate the relevant record in the internal database 118 indicating that the access request is suspect, but nonetheless allowing access to the secure application at step 414. This could also result in a case being created by generating and storing case record data in the management server as part of a case management system or two factor authentications can be requested.

If continuous occurrences of browser downgrades are found to be greater than 0.1% of all access requests received from all users over a period of time, an email alert action command (step 408) is executed to notify the administrator 116 that too many potential cases are being created and the pre-processing filter 202 executes an overload action command (step 416). This step allows the administrator to avoid overloading personnel that follow up fraud cases in the case management system. This can be an important step when new rules are being tested for the first time.

Each of the steps 404, 406, 408, 410 in the example process of Figure 4 is in the form of a process block. Steps 304 and 306 of the filter process may be represented as a series of process blocks arranged such that filter rules are applied to the access request data and resultant action commands are generated.

The filtering process 300 is performed by a rules engine 502 in the pre-processing filter 202 as shown in Figure 5. The rules engine 502 executes action commands relating to the customer devices 124 and the user device 102.

The access request data is received in the pre-processing filter 202 by an input adaptor 504 which translates the access request data from a variety of formats into a format recognised by the rules engine 502. For example, the input adaptor 504 can accept access requests for P \OPER\DBI "I9I II PCIr CoM p AU dC-II/)5/2()7 -9- Web services, http services and java APIs with the input being in a format corresponding to CSV, XML and/or a messaging system IBM MQ Series).

The rules engine 502 accesses the internal database 118 via a data connector 506 thus providing access to historical access request data and also has the ability to access data on the internal network during a user session with the secure application or via the Internet using the second data network 120. The rules engine 502 accesses and writes to the external database 122 via the second data network 120 using the same data connector 506 or a different data connector.

The specific arrangement or configuration of the rules engine 502 specific risk probability levels, and specific action commands) are selected by the administrator 116 using an editor 508 of the management server 112. The editor 508 is controlled by a user interface on the management console 114, a screen shot of which is shown in Figure 6. A further screen shot, shown in Figure 7, is a graphic representation of a plurality of process blocks which constitute the steps to be taken by the filter process 300 in an example configuration of the rules engine 502. The visual interface to the editor 508 advantageously allows rapid, convenient and error-free configuration and re-configuration of the particular filter process 300 performed by the pre-processing filter 202.

The process blocks which are available to be used in the rules engine 502 are stored in a rules catalogue 510. New rules may conveniently be updated from a third party provider of data security products over the Internet) or created ad-hoc by the administrator 116 using a process block creator in the editor 508. The set-up of the rules engine 502 is thus performed with an easy-to-use graphical interface and is highly flexible and adaptable to the needs of the owner 116 and the customer 126.

Example process blocks in the catalogue 510 include: Database access rules SMS alert rules IP address to location lookups P OPER\DBWV0I 1I 9 I rclops comp AU doc-1 Ilu/21x,7 Reverse DNS block list lookup rules Text manipulation rules The process blocks fall into one of three categories: 1. data analysis process blocks; (N 2. rule application process blocks; and 3. action command process blocks.

The data analysis process blocks extract data from the data submitted by the user, and perform manipulations of the data. For example, the data analysis process blocks may concatenate string data, access white or black-list data, retrieve historical data from the internal database 118, access geo-spatial data relating to an Internet Protocol address of the access request, generate data representing calculations of land speed deviations amounts etc, and generate analytical data based on patterns in the data submitted by the user (e.g.

click path, payment patterns).

The rule application process blocks control the process flow of the fraud detection filter.

For example, a rule application process block may compare data drawn from submissions by the user 104 to a constant value, or to data drawn from other submissions. A rule application process block may also execute policies in a loop, or in a sequence, or may exit a sequence.

The action command process blocks generate command data to be transmitted to external systems. For example, an action command process blocks may log selected data or add a case to a case management system. An action command process blocks may also generate alerts SMS, email) for the user 104 or the customer 126 or reject an access request.

A process block may also consist of a number of subsidiary process blocks linked so as to create a single process block.

P \OPE\DB%.II 1 I I opuuclOlp AU dc. I It5/21X7 -11- When creating and/or editing a series of process blocks to control processing of the rules engine 502, the administrator 116 may also test the processing of the arrangement in an off-line environment before running the new process in the rules engine 502) using a simulator 512. The simulator 512 allows the proposed filter process to be tested and observed in operation. The graphical user interface which displays the process blocks (e.g.

as shown in Figure 7) also graphically indicates the flow of the process during operation, thus enabling the administrator 116 to clearly visualise the operation of the proposed process.

The pre-processing filter 202 may be implemented using software code written in a computer program language, such as Java, running on a server engine JSP) and the application server 106 may be in the form of a server product such as J2EE. The management server 112 may be a J2EE server. Alternatively, the pre-processing filter 202 and management server 112 may be implemented at least in part by dedicated hardware circuits, such as ASICs and FPGAs, to further enhance processing speed, particularly if processing of the engine is not to be reconfigured regularly.

The logical implementation of the rules engine 502 is in the form of a multi-threaded design which provides high speed filtering. High speed filtering is used so that the user 104 does not notice an appreciable delay when accessing the secure application 204 via the pre-processing filter 202 if the access request is granted).

The external database 122 includes corporate databases, geospatial data, web services and black lists OFAC, SpamHaus, H-unter, Aristion, NetEconomy, and SearchSpace).

The pre-processing filter 202 and the management server 112 may be implemented on the same server as the secure application, or the management server 112 is separate as described above. The filter 202 is placed before the application 204 on the application server so as to provide access to the same session data and procedures as the secure application 204.

P NOPER\DBI)BVAOl I 91118 WopS Cop AU de- I I )5/2(X17 12- Many modifications will be apparent to those skilled in the art without departing from the scope of the present invention as herein described with reference to the accompanying drawings.

Claims (23)

1. A fraud detection filter installed in an application server including a secure application, said filter including: a rules engine for receiving request data representing an access request for the secure application from a user, and applying at least one risk condition rule to the request data for generating a risk probability level, and detecting at least one fraud condition when said risk probability level exceeds a threshold level, before passing said access request to said secure application.

2. A filter as claimed in claim 1, wherein said rules engine accesses past session data for said user in applying said at least one risk condition rule.

3. A filter as claimed in claim 1 or 2, wherein said rules engine accesses application data accessed by said application for said user in applying said at least one risk condition rule.

4. A filter as claimed in claim 3, wherein said application data for said user includes historical data. A filter as claimed in claim 3, wherein said application data includes account balance data.

6. A filter as claimed in any one of claims 1 to 5, including an input adaptor for receiving and processing said access request to provide said request data for said rules engine.

7. A filter as claimed in any one of claims 1 to 6, which invokes a two-factor authentication process for confirming the identity of the user when said fraud condition is detected. P \OPER\DBWA1 I oclopus cop AU doc- 11/)5/2-7

14- 8. A filter as claimed in claim 1, wherein applying at least one risk condition rule includes determining an Internet Protocol (IP) address associated with the access request. 9. A filter as claimed in claim 8, wherein applying at least one risk condition rule includes determining a change between the request data and previous request data representing a previous access request from the user. A filter as claimed in claim 9, wherein applying at least one risk condition rule includes determining a location associated with the IP address. 11. A filter as claimed in claim 10, wherein applying at least one risk condition rule includes determining a distance between the location and a previous location associated with a previous IP address associated with the previous access request. 12. A filter as claimed in claim 11, wherein applying at least one risk condition rule includes determining a speed from the distance, a receive time of the access request and a previous receive time of the previous access request. 13. A filter as claimed in any one of claims I and 8 to 12, wherein applying at least one risk condition includes determining a client parameter of a client application used to generate the access request. 14. A filter as claimed in claim 13, wherein applying at least one risk condition rule includes determining a change in the client parameter between the access request and a previous access request. A filter as claimed in either of claims 13 or 14, wherein the client parameter is the version of a Web browser used to generate the access request. P \OPIERUM)BM0) 191 I I KopOPS Cop AU do- 10/21,X7

16. A filter as claimed in any one of claims 1 and 8 to 15, wherein applying at least one risk condition rule includes determining a connection speed associated with the access request.

17. A filter as claimed in claim 16, wherein applying at least one risk condition rule further includes determining a speed change between the connection speed and a previous connection speed associated with a previous access request.

18. A filter as claimed in any one of claims 1 and 8 to 17, wherein applying at least one risk condition rule includes determining a connection type associated with the access request.

19. A filter as claimed in claim 18, wherein applying at least one risk condition rule includes determining a connection type change between the connection type and a previous connection type associated with a previous access request. A filter as claimed in either of claims 18 or 19, wherein determining the connection type includes determining when the access request is associated with a public hot- spot connection.

21. A filter as claimed in any one of claims 18 to 20, wherein determining the connection type includes determining when the access request is associated with a satellite connection.

22. A filter as claimed in claim 8, wherein applying at least one risk condition rule includes determining a blacklist match between the IP address and an 1P address blacklist.

23. A filter as claimed in any one of claims 1 to 22, wherein said risk probability level is generated using data produced by applying said at least one risk condition rule. P 'OPER\DBMIOu I I I IK ocopus comp AU doc-l I A /2(X)7 -16-

24. A filter as claimed in any one of claims 1 to 23, which denies access to the secure application for the user when said fraud condition is detected. A process as claimed in any one of claims 8 to 23, which invokes a two-factor authentication process for confirming the identity of the user when said fraud condition is detected, before passing said access request to said secure application.

26. A filter as claimed in any one of claims 1 to 24, including invoking an alert generation process for alerting a party when said fraud condition is detected, before passing said access request to said secure application.

27. A filter as claimed in claim 26, wherein the alert generation process includes generating an email alert or an SMS alert.

28. A filter as claimed in any one of claims I to 27, wherein applying at least one risk condition rule includes: generating a first risk probability level by applying a first risk condition rule to the request data; selecting a second risk condition rule based on the first risk probability level; and applying the second risk condition rule to the request data for generating a second risk probability level.

29. A management server for generating an interface to adjust and set said at least one risk condition rule as claimed in anyone of claims 1 to 28. A management server as claimed in claim 29, wherein said interface includes tools to adjust the dependence and connections between risk condition rules used to generate the risk probability level. P \OPER\DB'M')H1I I Ic opus comp AU dm-c I1 )5/2(Wl7 -17-

31. A filter system including a filter as claimed in any one of claims 1 to 28 and a management server as claimed in claim 29 or

32. An application server including: a secure application for access by a user; and a fraud detection filter as claimed in any one of claims 1 to 28.

33. A process for detecting a fraud condition, performed by an application server, including: receiving request data representing an access request from a user for a secure application of the application server; applying at least one risk condition rule to the request data for generating a risk probability level; and detecting said fraud condition when said risk probability level exceeds a threshold level, before granting access to said secure application.

34. A process as claimed in claim 33, including applying said at least one risk condition rule to subsequent access requests during a transaction session with said secure application before passing said access requests to said application. A process as claimed in claim 33, wherein said applying said at least one risk condition rule includes accessing past transaction session data for said user.