- Conference Article
- 10.1109/cyber-ai66431.2025.11233811
DarkWatchAI - Unsupervised Machine Learning for Identifying Fintech Threats in DarkWeb Posts
- Sep 01, 2025
- Luiza Nacshon + 3 more +3
Dark web posts are often cryptic and lengthy, presenting a significant challenge for identifying fintech cybersecurity threats, as threat actors use these forums to discuss attacks and sell illicit goods. To address this, we developed DarkWatchAI, an unsupervised machine learning framework for identifying and categorizing fintech threats in dark web content. Our novel methodology employs a transformed pointwise mutual information (T-PMI) model to cluster posts and identify correlations between fintech-related terms. These results are then fed via prompt engineering to a fine-tuned Generative AI (GenAI) LLM for threat actor categorization. This approach effectively extracts insights from complex dark web posts, identifying potential fintech security threats. Experiments show a 75% classification success rate in identifying cybersecurity threats. This work offers practical insights for dark web analysis, enabling financial institutions to proactively prevent fraud and protect customers against emerging threats.
Read more