Hi Nirmal, The data is not in hand right now.. but we can assume we have data.
Currently in Identity Server, it does not publish security related events such as login, logout, password change etc. What Asantha would do is write a Listener ( i.e extend AbstractUserOperationEventListener) and for each user operation it would publish an event. Once he does it, we can perform some user operations with a load test and gather data (get the events generated). Based on this data he needs to carry out the analysis. Thanks, TharinduE On Mon, Jun 27, 2016 at 6:22 PM, Nirmal Fernando <[email protected]> wrote: > Seshika implemented a fraud detection toolbox using Markov chain models > (using real-time siddhi queries) [1]. API-M Analytics product uses Markov > chain model to detect abnormal resource access patterns. > > In WSO2 ML, we have implemented a clustering based anomaly detection > algorithm [2]. > > [1] > http://wso2.com/analytics/solutions/fraud-and-anomaly-detection-solution/ > [2] > https://docs.wso2.com/display/ML110/Generating+a+Model+Using+the+K+Means+Anomaly+Detection+Algorithm+with+Labeled+Data > > > On Mon, Jun 27, 2016 at 5:45 PM, Tharindu Edirisinghe <[email protected]> > wrote: > >> +DamithN, Seshika, Nirmal >> >> >> @DamithN - I found the mail thread [1] bit similar. Are there any other >> reference to the work you've done for that ? >> >> @Seshika, Nirmal - Do you guys have any input for the work Asantha is >> trying to do ? He is a GSoC student that I mentor this year. >> >> Appreciate if you can help him out with this. >> >> >> [1] "[Architecture] Security Authentication Analytics" >> >> Thanks, >> TharinduE >> >> On Mon, Jun 27, 2016 at 4:15 PM, Asantha Thilina < >> [email protected]> wrote: >> >>> Hi all, >>> >>> i am a GSOC student who doing the *project 21 : NoSQL User Store >>> Development for Identity Server* and i am developing a *convolutional >>> neural network* to detect *frauds* using deeplearning4j[1] for my >>> research, i have done some case studies regarding fraud patterns ,i have >>> mainly focused on frauds that can be occur in online money transactions and >>> in login authentications i have refer some research papers[2],[3] and a >>> white paper[4] regrading those possible fraud patterns >>> I choosed convolutional neural network to develop my model but i am >>> little confused about how could include those fraud patterns in to my model >>> in a way it can detect a fraud in real time >>> >> > Once you have a built model, you could write a siddhi extension to perform > predictions (i.e. detect frauds in this case). > > Question: do we have data in hand? or are you trying to build a model on > the fly? > > and also is convolutional network is a best way to achieve my task or is >>> there any better method than this?,i would be grateful if anyone can guide >>> me to achieve this task >>> [1]http://deeplearning4j.org/convolutionalnets >>> [2]http://www.ijsce.org/attachments/File/NCAI2011/IJSCE_NCAI2011_025.pdf >>> [3] >>> https://www.researchgate.net/publication/200795976_Fraud_Detection_using_Neural_Networks >>> [4]https://neo4j.com/resources/fraud-detection-white-paper/ >>> >>> Thanks, >>> Asantha >>> >> >> >> >> -- >> >> Tharindu Edirisinghe >> Senior Software Engineer | WSO2 Inc >> Platform Security Team >> Blog : tharindue.blogspot.com >> mobile : +94 775181586 >> > > > > -- > > Thanks & regards, > Nirmal > > Team Lead - WSO2 Machine Learner > Associate Technical Lead - Data Technologies Team, WSO2 Inc. > Mobile: +94715779733 > Blog: http://nirmalfdo.blogspot.com/ > > > -- Tharindu Edirisinghe Senior Software Engineer | WSO2 Inc Platform Security Team Blog : tharindue.blogspot.com mobile : +94 775181586
_______________________________________________ Dev mailing list [email protected] http://wso2.org/cgi-bin/mailman/listinfo/dev
