Towards a Tricksy Group Shilling Attack Model against Recommender Systems

Wang, Youquan; Wu, Zhang; Cao, Jie; Fang, Changjian

doi:10.1007/978-3-642-35527-1_56

Cited by 11 publications

(8 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the shilling group generated by the group shilling attack model in Ref. [7] shows low similarity between group members, which causes DeR-TIA to fail. When detecting loosely coupled shilling groups on the Netflix dataset, the precision, recall, and F1-measure of CBS are 0.4221, 0.3758, and 0.3976, respectively.…”

Section: Comparison Of Detection Results On the Netflix Datasetmentioning

confidence: 99%

“…To generate the effective group shilling profiles and avoid the detection of the existing methods, Wang et al [7] proposed a tricky group shilling attack model which includes a strict version denoted as GSAGen s and a loose version denoted as GSAGen l . In their model, some shilling profiles generated by standard attack models, i.e., random attacks or average attacks, are used as the input.…”

Section: Background and Related Workmentioning

confidence: 99%

“…For the Netflix dataset, we have the ground truth information about the shilling groups because the injected shilling groups are generated by using group attack models proposed in Ref. [7]. For the sampled Amazon dataset, we have no ground truth information of shilling groups, e.g., the number of shilling groups, the size, the relationship between attack users and shilling groups, and so on.…”

Section: Evaluation Metricsmentioning

confidence: 99%

“…In fact, attack users can easily collude to attack a set of target items in the recommender system. e behaviour wherein a group of attack users collude to promote or demote the recommendation of a set of target items has been termed group shilling attacks [6,7]. As attack users in the group work together and strategically generate attack profiles, the detection methods proposed for detecting traditional attacks become ineffective for group shilling attacks.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

An Unsupervised Approach for Detecting Group Shilling Attacks in Recommender Systems Based on Topological Potential and Group Behaviour Features

Cai

Zhang

2021

Security and Communication Networks

View full text Add to dashboard Cite

To protect recommender systems against shilling attacks, a variety of detection methods have been proposed over the past decade. However, these methods focus mainly on individual features and rarely consider the lockstep behaviours among attack users, which suffer from low precision in detecting group shilling attacks. In this work, we propose a three-stage detection method based on strong lockstep behaviours among group members and group behaviour features for detecting group shilling attacks. First, we construct a weighted user relationship graph by combining direct and indirect collusive degrees between users. Second, we find all dense subgraphs in the user relationship graph to generate a set of suspicious groups by introducing a topological potential method. Finally, we use a clustering method to detect shilling groups by extracting group behaviour features. Extensive experiments on the Netflix and sampled Amazon review datasets show that the proposed approach is effective for detecting group shilling attacks in recommender systems, and the F1-measure on two datasets can reach over 99 percent and 76 percent, respectively.

show abstract

Section: Comparison Of Detection Results On the Netflix Datasetmentioning

confidence: 99%

Section: Background and Related Workmentioning

confidence: 99%

Section: Evaluation Metricsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

An Unsupervised Approach for Detecting Group Shilling Attacks in Recommender Systems Based on Topological Potential and Group Behaviour Features

Cai

Zhang

2021

Security and Communication Networks

View full text Add to dashboard Cite

show abstract

“…Traditional attack models are based on statistics methods, such as random (Ran) attack model, average (Avg) attack model, average over popular (AoP) items attack model, and power item attack (PIA) model. Deep learning-based attack models use deep learning methods to construct shilling profiles by simulating rating profiles of genuine users, such as Graph cOnvolution based generative shilling ATtack (GOAT) model [5].…”

Section: Background and Related Workmentioning

confidence: 99%

A Shilling Group Detection Framework Based on Deep Learning Techniques

Wang

Zhang

Wang

et al. 2022

Security and Communication Networks

View full text Add to dashboard Cite

Research studies have shown that online recommender systems are under the threat of group shilling attacks, in which attackers attempt to distort the recommendation results of particular items by cooperatively injecting fake profiles. Existing detection methods usually divide candidate groups at first and then use the hand-crafted features to recognize shilling groups. However, the detection performance of existing methods depends highly on the quality of candidate groups obtained. Moreover, extracting features manually is time-consuming. To overcome these limitations, we propose a shilling group detection framework based on the sparse autoencoder and modified GraphSAGE model. First, we use the sparse autoencoder to obtain rating features of users from the rating dataset. Second, we analyse user collusive degrees to calculate user transition probabilities. Third, we build a user relation graph and utilize the modified GraphSAGE model to perform user classification. Finally, shilling groups are gathered according to the neighbour relations. Extensive experiment demonstrates that the proposed framework performs better on different datasets than baseline methods.

show abstract