An Attention-Based Graph Neural Network for Spam Bot Detection in Social Networks

Abstract

With the rapid development of social networks, spam bots and other anomaly accounts’ malicious behavior has become a critical information security problem threatening the social network platform. In order to reduce this threat, the existing research mainly uses feature-based detection or propagation-based detection, and it applies machine learning or graph mining algorithms to identify anomaly accounts in social networks. However, with the development of technology, spam bots are becoming more advanced, and identifying bots is still an open challenge. This paper proposes a new semi-supervised graph embedding model based on a graph attention network for spam bot detection in social networks. This approach constructs a detection model by aggregating features and neighbor relationships, and learns a complex method to integrate the different neighborhood relationships between nodes to operate the directed social graph. The new model can identify spam bots by capturing user features and two different relationships among users in social networks. We compare our method with other methods on real-world social network datasets, and the experimental results show that our proposed model achieves a significant and consistent improvement.

Visual analysis of global research trends in social bots based on bibliometrics

“…According to the feature analysis, the detection methods (Adewole et al , 2017) can be divided into: (1) methods based on social network analysis, involving analyzing the topological social structure of the accounts within the network or extracting discriminative network features to detect social bots (Mendoza et al , 2020; Lingam et al. , 2019; Zhao et al , 2020); (2) methods based on content and behavioral analysis, involving distinguishing social bots from humans through profile information, textual features, URL features, topic/mention/retweet features, posting time features, sentiment features, etc. (Cai et al.…”

“…According to the detection technologies, the existing social bot detection methods include (Orabi et al, 2020): (1) graph-based methods (Ahmed and Abulaish, 2013;Dorri et al, 2018); (2) machine learning-based methods, including the approach of supervised learning (Alarifi et al, 2016;Chu et al, 2012;Kudugunta and Ferrara, 2018), unsupervised learning (Ahmed and Abulaish, 2013;Cresci et al, 2018) and semi-supervised learning (Dorri et al, 2018;Shi et al, 2019); (3) crowdsourcing-based methods (Wang et al, 2012;Cresci et al, 2017); and (4) anomaly-based methods (Wang and Paschalidis, 2017;Costa et al, 2017;Pan et al, 2016). According to the feature analysis, the detection methods (Adewole et al, 2017) can be divided into: (1) methods based on social network analysis, involving analyzing the topological social structure of the accounts within the network or extracting discriminative network features to detect social bots (Mendoza et al, 2020;Lingam et al, 2019;Zhao et al, 2020); (2) methods based on content and behavioral analysis, involving distinguishing social bots from humans through profile information, textual features, URL features, topic/mention/retweet features, posting time features, sentiment features, etc. (Cai et al, 2017;Liu, 2019;Rout et al, 2020); (3) methods based on hybrid analysis, that is the methods combining both content/ behavioral and network information at the same time (Wang, 2010;Dorri et al, 2018;Fazil and Abulaish, 2020).…”

Hybrid-MELAu: A Hybrid Mixing Engineered Linguistic Features Based on Autoencoder for Social Bot Detection.

“…In addition to these works based on supervised and unsupervised approaches, present studies utilize a semisupervised method to identify social bots. Zhao et al [30] present a semi-supervised model founded on a attention mechanism-based graph CNN, which spots spam bots by integrating many user characteristics and relational structures. To detect counterfeit accounts from a vast volume of Twitter data, BalaAnand et al [31] presented an enhanced graph-based semi-supervised learning algorithm (EGSLA).…”

Adversarial Machine Learning on Social Network: A Survey

“…Zhao et al [68] introduced a semi-supervised graph embedding model to detect spam bot for the directed social network, where they used the attention mechanism and graph neural network to detect spam bot based on the retweet relationship and the following relationship between users. They experimented with the Twitter 1KS-10KN dataset [69] which was collected on Twitter, compared with GCN, GraphSAGE, and GAT, their method achieved the best performance in Recall, Precision, and F1-score.…”