Fraud Detection with Neo4j

FraudDemoSolu%onspoweredwithNeo4j

2015StefanKolmarNeoTechnologies

RetailBankingFirst-PartyFraud

Openingmanylinesofcreditwithnointen5onofpayingthemback

CausingHighImpact•  TensofbillionsofdollarslosteveryyearbyU.S.Banks.(1)

•  25%oftotalconsumercreditcharge-offsintheUnitedStates.(2)

•  10%to20%ofunsecuredbaddebtatleadingU.S.andEuropeanbanksismisclassified,andisactuallyfirst-partyfraud.(3)

(1)Experian:hQp://www.experian.com/assets/decision-analy5cs/white-papers/first-partyfraud-wp.pdf(2)Experian:hQp://www.experian.com/assets/decision-analy5cs/white-papers/first-partyfraud-wp.pdf(3)BusinessInsider:hQp://www.businessinsider.com/how-to-use-social-networks-in-the-fight-against-first-party-fraud-2011-3

Detec%ngFraudRings

SSN1

123NW1stStreetSanFrancisco,CA

555-555-5555

123NW1stStreetSanFrancisco,CA555-5

55-5555

Skimming

PersonA PersonB

Loca5onA Loca5onB

PhoneNumberDuplicateUse

555-555-5555

PersonA

PersonB

SuspecteCommerce

PersonA

PersonB

Loca5onC IPaddress

FraudDemo–PartI(generic)•  FraudscenariocoveringRetailFraudusecases•  Datasetcontainsopera5onaldata•  Constantdataload–>injec5ngfraudcases->generatealerts•  Capabilitytoexportdataofdetectedfraudforfurtherinves5ga5on

Neo4j

AppServer

FraudDetec5onWebApp

FraudAppBrowser

UX:TestDataGen

Alertgenerated

Demo

WhyusingGraphDB/Neo4jforFraud?

•  AgileDevelopment•  Highproduc5vityandrapidimplementa5on•  No“RDBMS-waterfall-high-investment-trap”

•  Takingadvantageofthefullvalueofconnecteddata•  TraversingthegraphcomparedtoselfjoinsinRDBMS

•  Nearreal5meresponse5mes•  Preven5ngfraudratherthandetec5ngaderthefact

•  Schemafree•  Nodescanvarydependingon5me/usage/seman5c

•  UsagescenarioFraudAnalyst:•  Poten5alfraudcasedetected•  Enrichedwithdatafromvarioussourcescontainingdataonfraudsuspect•  Triggerhumanand/orautomatedreac5ons

FraudDemo–PartII

Neo4j

WebApp Data

Integra5on

RDBMS(Oracle,MySQL,DB2,HANA…)

ManagementConsole(E.gBIToolssuchasTableau,Qlik,BO,MicroStrategyetc)

FraudAnalyst

Machine2Machinegeneratedac5ons

Alert

IncomingEvents

CRMSystem

Opera5onalSystem

ExternalData

UsingNeoasthefounda%onofafraudsolu%oninyourarchitecture

Step1:SetupDataIntegra5onStep2:VisualizeDatainBITool

Conclusions

•  Fraudasoneusecasetoprovidefullvalueofconnecteddatawithintheen5reorganiza5on

•  Neo4jasthefounda5ontodo360degreefrauddetec5onandpreven5on

•  Neo4jtoextendyourexis5ngenvironmentwhileprotec5ngyourinvestments

•  Neo4jprovidesbestvalueintegratedintheen5reenvironment•  Neo4jasthefounda5onforgenera5ngreal5mealertstotriggerautomatedormanualinterven5ons

Encore

…Onemorething...

Abrieflookintothedatamodel….

FraudDemoSolu%onspoweredwithNeo4j

[email protected]