System for Integrating Social Network Analysis and Information Security in the Context of Countering Radicalism

Abstract

The spread of destructive and radical content on social media is a growing threat to national and global security. Existing countermeasures often focus only on single aspects (text or network structure) and prove ineffective against veiled propaganda and dynamic information dissemination. This paper proposes the development of a comprehensive system for social media content evaluation that integrates multimodal analysis, social network analytics (SNA) techniques and temporal dynamics. The proposed system analyzes the visual content of posts using neural networks (YOLO), evaluates textual content for relevance to target topics using natural language processing and classification techniques (e.g., BERT), represents social relationships as a graph for influence analysis, and introduces a freshness metric to account for content relevance. The results of each component analysis are integrated into a single final post score, based on which the post is classified into relevance/influence. The key metrics and methods applied at each stage are discussed, and an example of how the system works in the context of identifying potentially radical content is demonstrated. The presented system provides more accurate and timely identification of destructive publications and understanding of their distribution patterns, which is critical for effectively countering radicalization in the digital space. The study was funded by the Russian Science Foundation (RNF) grant (project No. 25-21-20125 “Development of mathematical models and a system for assessing content for extremist content based on text analysis and object detection in images and videos”).