Research Article
A Sybil Detection Method in OSN Based on DistilBERT and Double-SN-LSTM for Text Analysis
@INPROCEEDINGS{10.1007/978-3-030-90022-9_4, author={Xiaojie Xu and Jian Dong and Zhengyu Liu and Jin Yang and Bin Wang and Zhaoyuan Wang}, title={A Sybil Detection Method in OSN Based on DistilBERT and Double-SN-LSTM for Text Analysis}, proceedings={Security and Privacy in Communication Networks. 17th EAI International Conference, SecureComm 2021, Virtual Event, September 6--9, 2021, Proceedings, Part II}, proceedings_a={SECURECOMM PART 2}, year={2021}, month={11}, keywords={Sybil Attack DistilBERT Double-SN-LSTM}, doi={10.1007/978-3-030-90022-9_4} }- Xiaojie Xu
Jian Dong
Zhengyu Liu
Jin Yang
Bin Wang
Zhaoyuan Wang
Year: 2021
A Sybil Detection Method in OSN Based on DistilBERT and Double-SN-LSTM for Text Analysis
SECURECOMM PART 2
Springer
DOI: 10.1007/978-3-030-90022-9_4
Abstract
Sybil attacks are increasingly rampant in online social networks (OSNs); thus, Sybil detection is one of the key issues in OSN security research. Sybils in OSNs are often used by attackers for public opinion intervention, topic flow filling, and dissemination of false and malicious messages. Therefore, if the credibility of the Sybil can be analyzed, then the harm of Sybil attacks can be prevented to a certain extent. Based on the analysis of existing Sybil detection research, this paper proposes an end-to-end Sybil detection model based on the Bidirectional Encoder Representations from Transformers (BERT) model that analyzes tweet text content. Considering the problems of the existing datasets, we built a dataset for text content analysis of tweets based on the hot political topic of the 2020 US presidential election. Accordingly, this study used a distilled version of BERT, DistilBERT, as the sentence embedding model, and the double self-normalizing long short-term memory (Double-SN-LSTM) recurrent neural network model as the classification detection model. The final experimental effect was greatly improved compared with the existing analysis methods, and it had a better detection effect for the more concealed Sybils.
