Blockchain-Enhanced TLS Session Metadata Classification Using Machine Learning for Secure and Auditable Traffic Analysis

Transport Layer Security (TLS) encryption secures internet communications but obscures malicious traffic, compli- cating traditional detection methods. This paper proposes an in- novative framework that integrates blockchain technology, AES- CBC encryption, and machine learning to securely store, enrich, and classify TLS session metadata. Flow- level features, extracted from passive network captures, are encrypted and immutably logged on a private blockchain, ensuring confidentiality and auditability. A decision tree classifier, trained offline on decrypted metadata, achieves 93.2% accuracy, 92.8% precision, and 91.6% recall in distinguishing benign from malicious sessions. The system’s modular architecture supports scalability and lays the foundation for real-time intelligent firewalls. Experimental results on a 10,000- session dataset validate the approach, demonstrating superior performance compared to baseline methods and poten- tial for enterprise-grade deployment.