Research & Papers

SiNMULI achieves 99.89% accuracy detecting malicious URLs

New signed network approach detects phishing URLs with near-perfect accuracy using social network analysis

Deep Dive

Researchers Avijit Gayen, Sayan Mondal, and Angshuman Jana have proposed SiNMULI (Signed Network Approach for Malicious URL Identification), a groundbreaking method for detecting malicious websites using signed network analysis.

The approach models URL relationships as a graph where nodes represent domains and edges represent backlinks with polarity (trusted/untrusted). Using balance-theoretic inference, the system propagates edge signs through the network and classifies unlabeled domains based on a 51% majority rule of incoming links. When tested on real-world datasets, SiNMULI achieved 99.89% accuracy, 99.62% precision, and 99.80% F1-score - outperforming conventional machine learning and deep learning approaches. The method's interpretability, adversarial resilience, and independence from training data make it particularly suitable for real-time cyber defense applications.

Key Points
  • SiNMULI achieves 99.89% accuracy detecting malicious URLs using signed network analysis
  • The method uses balance-theoretic inference on URL backlink graphs with polarity (trusted/untrusted edges)
  • Outperforms traditional ML/deep learning models while being lightweight and resilient to adversarial attacks

Why It Matters

Near-perfect malicious URL detection could revolutionize cybersecurity defenses against phishing and fraud

📬 Get the top 10 AI stories daily