BankSealer : An Online Banking Fraud Analysis and Decision Support System
We propose a semi-supervised online banking fraud analysis and decision support approach. During a training phase, it builds a profile for each customer based on past transactions. At runtime, it supports the analyst by ranking unforeseen transactions tha
- PDF / 1,028,033 Bytes
- 15 Pages / 439.363 x 666.131 pts Page_size
- 21 Downloads / 233 Views
Politecnico di Milano, Italy Dipartimento di Elettronica, Informazione e Bioingegneria {michele.carminati,roberto.caron,federico.maggi,stefano.zanero}@polimi.it 2 Politecnico di Milano, Italy Dipartimento di Matematica [email protected] Abstract. We propose a semi-supervised online banking fraud analysis and decision support approach. During a training phase, it builds a profile for each customer based on past transactions. At runtime, it supports the analyst by ranking unforeseen transactions that deviate from the learned profiles. It uses methods whose output has a immediate statistical meaning that provide the analyst with an easy-to-understand model of each customer’s spending habits. First, we quantify the anomaly of each transaction with respect to the customer historical profile. Second, we find global clusters of customers with similar spending habits. Third, we use a temporal threshold system that measures the anomaly of the current spending pattern of each customer, with respect to his or her past spending behavior. As a result, we mitigate the undertraining due to the lack of historical data for building of well-trained profiles (of fresh users), and the users that change their (spending) habits over time. Our evaluation on real-world data shows that our approach correctly ranks complex frauds as “top priority”. Keywords: fraud detection, bank fraud, anomaly detection.
1
Introduction
The popularity of Internet banking has led to an increase of frauds, resulting in substantial financial losses [15,4]. Banking frauds increased 93% in 2009–2010 [6], and 30% in 2012–2013 [8]. Internet banking frauds are difficult to analyze and detect because the fraudulent behavior is dynamic, spread across different customer’s profiles, and dispersed in large and highly imbalanced datasets (e.g., web logs, transaction logs, spending profiles). Moreover, customers rarely check their online banking history such regularly that they are able to discover fraud transactions timely [15]. We notice that most of the existing approaches build black box models that are not very useful in manual investigation, making the process slower. In addition, those based on baseline profiling are not adaptive, also due to cultural and behavioral differences that vary from country to country. Instead of focusing on N. Cuppens-Boulahia et al. (Eds.): SEC 2014, IFIP AICT 428, pp. 380–394, 2014. c IFIP International Federation for Information Processing 2014
BankSealer
381
pure detection approaches, we believe that more research efforts are needed toward systems that support the investigation, and we had the unique opportunity to work on a real-world, anonymized dataset. In this paper we propose BankSealer, an effective online banking semisupervised decision support and fraud analysis system. BankSealer automatically ranks frauds and anomalies in bank transfer transactions and prepaid phone and debit cards transactions. During a training phase, it builds a local, global, and temporal profile for each user. The local profile models past user behavior
Data Loading...
