Secure Ensemblechain for Decentralized and Robust Machine Learning

Abstract

The rise of ensemble learning has led to significant advancements in machine learning, providing improved accuracy and generalization by combining multiple models. However, the centralized nature of traditional ensemble systems introduces vulnerabilities, such as single points of failure and exposure to malicious attacks. To address these challenges, we introduce secure EnsembleChain, a novel decentralized system that leverages blockchain technology to enhance the security, robustness, and trustworthiness of ensemble learning. By leveraging blockchain's distributed ledger and consensus mechanisms, secure EnsembleChain mitigates risks associated with centralized systems, ensuring trust and security in model collaboration. Smart contracts automate malicious node detection, while an immutable record fosters transparency. In this paper, we focus only on the bagging ensemble technique. Experimental results show that secure EnsembleChain improves resilience against attacks and offers an efficient, scalable solution for decentralized AI collaboration, combining the strengths of AI and the blockchain technology.