Clustering and Labeling Auction Fraud Data

Abstract

Although shill bidding is a common auction fraud, it is however very tough to detect. Due to the unavailability and lack of training data, in this study, we build a high-quality labeled shill bidding dataset based on recently collected auctions from eBay. Labeling shill biding instances with multidimensional features is a critical phase for the fraud classification task. For this purpose, we introduce a new approach to systematically label the fraud data with the help of the hierarchical clustering CURE that returns remarkable results as illustrated in the experiments.

Incremental learning framework for real‐world fraud detection environment

“…To scale up with the rapid and continuous flow of auction data, we utilize multiple servers to produce the Shill Bidding chunks. The previous study 10 described the Shill Bidding data generated from commercial auctions. In Section 5, we explain the Shill Bidding patterns and data.…”

“…Each Shill Bidding feature describes a unique bidding behavior. 9,10 A good quality Shill Bidding dataset was developed in Reference 10 using commercial auctions of eBay and then labeled in Reference 9 using a validated hierarchical clustering method. Table 3 provides some statistics about the Shill Bidding dataset.…”

“…A jewelry seller was accused of Shill Bidding on eBay in 2007 and paid $400,000 for the settlement. 10 Besides, the seller and his staff were prevented from participating in eBay for 4 years. Consequently, it is vital to examine bidders' behavior to track Shill Bidding, hence preventing innocent buyers from being overcharged.…”

“…• Scarcity of labeled data: Annotating multidimensional data comes with a high cost as it is usually done manually by the experts of the application domain. In the Shill Bidding detection context, this task consisted mainly of two time-consuming operations: 10 first, optimally cluster the Shill Bidding data into groups with similar bidding behaviors, and second, label manually each group according to the values and weights of the Shill Bidding features. • Relevancy of incremental data: It is necessary to determine which data are most adequate to the gradual adaptation of the Shill Bidding classification model.…”

Building High-Quality Auction Fraud Dataset

“…To the best of our knowledge, this will be the first attempt to employ SCC for SB detection. For this purpose, we need first to label a portion of our SB dataset, for instance through data clustering techniques (Alzahrani & Sadaoui, 2019). Moreover, we will study how to deal with the problem of imbalanced class distribution, which is expected in fraud classification applications (Ganguly & Sadaoui, 2017); (Anowar, Sadaoui & Mouhoub, 2018).…”