Blockchain-Assisted Federated Learning for Cybersecurity: Combining Isolation Forest, Variational Autoencoders, and Differential Privacy
Durga Praveen Deevi, Naga Sushma Allur, Koteswararao Dondapati, Himabindu Chetlapalli, Sharadha Kodadi, Aravindhan Kurunthachalam · Journal of Science and Technology · 2025
The complexity of the cyber threats dictates the need for strong, privacy-preservingmechanisms for anomaly detection. This paper introduces a new framework called BAFL, anintegration of Isolation Forest and Variational Autoencoders combined with DifferentialPrivacy, for safe and scalable solutions in cybersecurity applications. Federated Learningallows distributed training across numerous clients without exposure of sensitive information,while the blockchain technology introduces trust and integrity in model updates.