iBGP: A Bipartite Graph Propagation Approach for Mobile Advertising Fraud Detection

Abstract

Online mobile advertising plays a vital financial role in supporting free mobile apps, but detecting malicious apps publishers who generate fraudulent actions on the advertisements hosted on their apps is difficult, since fraudulent traffic often mimics behaviors of legitimate users and evolves rapidly. In this paper, we propose a novel bipartite graph-based propagation approach, iBGP, for mobile apps advertising fraud detection in large advertising system. We exploit the characteristics of mobile advertising user’s behavior and identify two persistent patterns: power law distribution and pertinence and propose an automatic initial score learning algorithm to formulate both concepts to learn the initial scores of non-seed nodes. We propose a weighted graph propagation algorithm to propagate the scores of all nodes in the user-app bipartite graphs until convergence. To extend our approach for large-scale settings, we decompose the objective function of the initial score learning model into separate one-dimensional problems and parallelize the whole approach on an Apache Spark cluster. iBGP was applied on a large synthetic dataset and a large real-world mobile advertising dataset; experiment results demonstrate that iBGP significantly outperforms other popular graph-based propagation methods.

Detecting problematic transactions in a consumer-to-consumer e-commerce network

“…Another reason for which we focused on local properties was that we could not obtain the global network structure for computational reasons. It should be noted that, while the use of global network properties in addition to local ones may improve the classification accuracy (Bhat and Abulaish 2013), the present local method attained a similar classification performance to those based on global network properties, i.e., 0.880-0.986 in terms of the ROC AUC (Šubelj et al 2011;Van Vlasselaer et al 2015;Van Vlasselaer et al 2016;Hu et al 2017;Li et al 2017;Savage et al 2017). A prior study using data from the same marketplace, Mercari, aimed to distinguish between desirable non-professional frequent sellers and undesirable professional sellers (Yamamoto et al 2019).…”

“…Prior network-based fraud detection has employed either global or local network properties to characterize nodes. Global network properties refer to those that require the structure of the entire network for calculating a quantity for individual nodes, such as the connected component (Šubelj et al 2011;Savage et al 2017;Wang et al 2018), betweenness centrality (Šubelj et al 2011;Dreżewski et al 2015 Akoglu et al 2013;Bangcharoensap et al 2015;Van Vlasselaer et al 2015Li et al 2017;Hu et al 2017), dense subgraphs including the case of communities (Šubelj et al 2011;Bhat and Abulaish 2013;Ferrara et al 2014;Jiang et al 2014;Hooi et al 2016;Liu et al 2016;Shchur et al 2018), and k-core (Wang and Chiu 2008;Rasheed et al 2018). Although many of these methods have accrued a high classification performance, they require the information about the entire network.…”

“…Note that we analyze transaction frauds based on transaction networks of users, which contrasts with previous studies of online C2C marketplaces that looked at reputation frauds (Chau et al 2006;Pandit et al 2007;Wang and Chiu 2008;Yanchun et al 2011). Many prior network-based fraud detection algorithms used global information about networks, such as connected components, communities, betweenness, k-cores, and that determined by belief propagation (Chau et al 2006;Pandit et al 2007;Wang and Chiu 2008;Šubelj et al 2011;Akoglu et al 2013;Bhat and Abulaish 2013;Ferrara et al 2014;Jiang et al 2014;Bangcharoensap et al 2015;Dreżewski et al 2015;Van Vlasselaer et al 2015;Hooi et al 2016;Liu et al 2016;Van Vlasselaer et al 2016;Colladon and Remondi 2017;Hu et al 2017;Li et al 2017;Liu et al 2017;Savage et al 2017;Shchur et al 2018;Rasheed et al 2018;Wang et al 2018). Others used local information about the users' network, such as the degree, the number of triangles, and the local clustering coefficient (Chau et al 2006;Akoglu et al 2010;Šubelj et al 2011;Yanchun et al 2011;Bhat and Abulaish 2013;Bangcharoensap et al 2015;Dreżewski et al 2015;Monamo et al 2016;Van Vlasselaer et al 2016;Colladon and Remondi 2017).…”

“…These methods stand on the expectation that anomalous users show connectivity patterns that are distinct from those of normal users (Akoglu et al 2015). Network analysis has been deployed for fraud detection in insurance (Šubelj et al 2011), money laundering (Dreżewski et al 2015;Colladon and Remondi 2017;Savage et al 2017), health-care data (Liu et al 2016), carbooking (Shchur et al 2018), a social security system (Van Vlasselaer et al 2016), mobile advertising (Hu et al 2017), a mobile phone network (Ferrara et al 2014), online social networks (Bhat and Abulaish 2013;Jiang et al 2014;Hooi et al 2016;Rasheed et al 2018), online review forums (Akoglu et al 2013;Liu et al 2017;Wang et al 2018), online auction or marketplaces (Chau et al 2006;Pandit et al 2007;Wang and Chiu 2008;Bangcharoensap et al 2015;Yanchun et al 2011), credit card transactions (Van Vlasselaer et al 2015;Li et al 2017), cryptocurrency transaction (Monamo et al 2016), and various other fields (Akoglu et al 2010). For example, fraudulent users and their accomplices were shown to form approximately bipartite cores in a network of users to inflate their reputations in an online auction system (Chau et al 2006).…”

GFD: A Weighted Heterogeneous Graph Embedding Based Approach for Fraud Detection in Mobile Advertising

“…What is more, fraudsters could easily adjust their fraud patterns based on existing fraud detection attributes and rules to avoid being detected. Recently, some researchers try to use the relationship between information entities to construct a graph model and then use the graph mining or learning methods to identify the changing fraud behaviors [15][16][17]. All these methods obtain useful insights into the learning mechanism to classify fraud behaviors from normal activities.…”