Cooperative Fraud Detection Model With Privacy-Preserving in Real CDR Datasets

Abstract

The researchers have shown broad concern about detection and recognition of fraudsters since telecommunication operators and the individual user are both suffering significant losses from fraud activities. Researchers have proposed various solutions to counter fraudulent activity. However, those methods may lose effectiveness in fraud detection because fraudsters always tend to cover their tracks by roaming among different telecommunication operators. What is more, due to the lack of real data, researchers have to do simulations in a virtual scenario, which makes their models and results less persuasive. In our previous paper, we proposed a novel strategy with high accuracy and security through cooperation among mobile telecommunication operators. In this manuscript, we will validate it in a real-world scenario using real Call Detail Records(CDR) data. We apply the Latent Dirichlet Allocation (LDA) model to profile users. Then we use a method based on Maximum Mean Discrepancy (MMD) to compare the distribution of samples to match roaming fraudsters. Cooperation between telecommunication operators may boost the accuracy of detection while the potential risk of privacy leakage exists. A strategy based on Differential Privacy(DP) is used to address this problem. We demonstrate that it can detect the fraudsters without revealing private data. Our model was validated using simulated dataset and showed its effectiveness. In this manuscript, experiments are performed on real CDRs data, and the result shows that our method has impressive performance in the real-world scenario as well.

Insight into Anomaly Detection and Prediction and Mobile Network Security Enhancement Leveraging K-Means Clustering on Call Detail Records

“… Fraud Detection and Prevention: CDR data analysis is instrumental in detecting and preventing telecommunications fraud. By analyzing usage patterns and detecting anomalies in call behavior, operators can identify fraudulent activities such as call spoofing, SIM box fraud, and premium rate service fraud [ 4 ]. Customer Experience Management: CDR data analysis enables telecommunications operators to better understand customer behavior and preferences.…”

Money Transaction Fraud Detection Using Harris Grey Wolf-Based Deep Stacked Auto Encoder

“…AccordingtotheAmericanheritagedictionary,fraudistermedasthedeceptionthatisintentionally accomplishedinordertoperformtheunlawfulgain.However,frauddetectionisusedtoidentify thesymptomsoffraud.Someoftheexamplesthatincludefrauddetectionstrategiesareaccounting fraud, credit card fraud, and insurance fraud. However, the information collected from Nigeria Inter-bank settlement system (NIBSS) showed that the fraudulent transactions are at peak levels inthebankingsector.Inrecentdecades,thefraudsarebeingevolvedandcommittedfromcasual fraudstersintotheorganizedcrimesuchthatthefraudringsutilizedsomesophisticatedtechniques tocontrolthecommitandaccountsoffraud.AccordingtotheresearchofJavelinin2007,nearly 6.8millionAmericansareoffendedthroughcardfraud.In2007,thefraudsintheexistingaccounts faced$3billionloss.Moreover,theNilsonreportspecifiesthecostoftheindustryas$4.84billion (Makki,et al,2019;Nilson,2007;John,et al,2016).In2007,Javelinstatesthetransactionloss amountas%30.6billion (Kim,&Monathan,2008;John,et al,2016).In2006,cardfraudmakes thetransactionlossof423millionaccordingtotheUKeconomy.Moreover,creditcardfraudloses accounttransactionsof$600milliongloballyeachyear (John,et al,2016).Fraudisthecriminalor wrongfuldeceptionthatisintendedtobringthepersonalorthefinancialgain (Zanetti,et al,2017;John,et al,2016;Randhawa,et al,2018).Theresearchworksinfraudreductionarecategorizedinto twotopics(i)frauddetectionand(ii)fraudprevention.Fraudpreventionistheproactiveapproach, wherethecauseoffraudisstopped (Gowthami,et al,2018;Cristin,et al,2019).However,fraud detectionishighlyrequiredwhenfraudulenttransactionsareacceptedbythefraudster (Ruan,et al, 2019;Randhawa,et al,2018).…”