Algerian Journal of Forensic Sciences and Criminology
Volume 2, Numéro 2, Pages 17-24
2025-12-29

A New Hybrid Sna-xai Framework For Detecting And Explaining Disinformation In Criminal Social Networks

Authors : Jouini Abdelhafid . Lebbaz Noureddine .

Abstract

The spread of disinformation across digital communication networks poses a significant challenge to law enforcement and national security, particularly when exploited by criminal organizations to coordinate activities, manipulate public opinion, and generate large-scale disruption. This paper introduces an innovative hybrid SNA-XAI framework that integrates Social Network Analysis (SNA) with eXplainable Artificial Intelligence (XAI) to detect and explain disinformation within criminal social networks The proposed framework extends epidemiological modeling, specifically the Susceptible-Infected-Recovered (SIR) model, and employs SNA techniques to characterize network structures that facilitate the rapid diffusion of disinformation and to identify influential actors. It further incorporates XAI methods, including Shapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME), to generate transparent linguistic evidence, elucidate disinformation content, and interpret the mechanisms underlying its propagation. Through a simulated real-world case study of a coordinated disinformation campaign, the framework demonstrates strong performance in identifying both key actors and disinformation content while maintaining high levels of transparency and interpretability. The integration of structural and content-based analyses enables rapid, evidence-driven alerts that assist security analysts in mitigating emerging threats. Experimental results confirm that the framework substantially enhances detection speed, accuracy, and explainability, thereby supporting its potential for real-world deployment and future integration with multimedia analysis.

Keywords

Disinformation, Criminal digital networks, Explainable AI (XAI), Social Network Analysis (SNA).