SN-Stego: Dataset for Social Networks Text Steganalysis via Local Group Discovery and Sample Distribution Regulation
Qiong Xu, Ru Zhang, Jianyi Liu, Yongfeng Huang · Data Intelligence · 2025
Social networks’ rapid information dissemination, massive user bases, and diverse content make them vulnerable to text steganography—a covert technique embedding secret messages into texts undetected, threatening personal privacy and network security. While text steganalysis serves as a critical defense mechanism, existing datasets for this task suffer from critical limitations including missing social graphs, insufficient text attributes, mismatched sample distributions, and limited data scale, hindering research progress. To address these gaps, this paper proposes a novel methodology for constructing a social network text steganalysis dataset via meta pathconstrained local group discovery and sample distribution dynamic regulation. It utilizes a local group discovery algorithm constrained by “user-tweet-hashtag” meta path to sample special user groups with potential covert communication intentions. In addition, a three-dimensional dynamic regulation strategy is designed to reshape the original tweets of the special users by adjusting the ratio, type, and distribution of steganographic texts, simulating complex and diverse covert communication patterns. Finally, a dataset is constructed with rich social graph information, namely SN-stego. It conforms to the characteristics of text fragmentation and steganography sparsity in real social networks, and simulates various social network text steganography analysis scenarios with complex and diverse sample distributions. Statistical analyses and empirical evaluations demonstrate that SN-stego exhibits substantial advancements in data scale, entity diversity, and scenario adaptability. The proposed method provides solid technical support for expanding and deepening the research on text steganalysis in social networks.