FNDEX: Fake News and Doxxing Detection with Explainable AI

Fake news and doxxing often work together as coordinated threats: The paper identifies that fake news and doxxing—the malicious sharing of private information—frequently intersect online, yet this dangerous combination had not been systematically studied before FNDEX was developed.

AI systems need to explain their decisions to be trustworthy: The research emphasizes that detecting harmful content isn't enough; the system must provide clear explanations for why it flagged something as fake news or doxxing so users and moderators can understand and verify the AI's reasoning.

Privacy protection is essential when building detection systems: FNDEX uses a three-step anonymization process to protect personally identifiable information while developing the detection models, demonstrating that responsible AI development requires safeguarding data even during the research phase.

Specialized AI models outperform generic approaches: The system uses three distinct transformer models working together to detect both fake news and doxxing, significantly outperforming existing baseline systems, suggesting that targeted AI solutions are more effective than one-size-fits-all approaches.

The challenge of balancing free speech with online safety is urgent: The paper frames fake news and doxxing as emerging "cyber weapons" that exploit the tension between freedom of expression and platform safety, highlighting why automated detection tools have become necessary for responsible internet governance.