Turan Jafar-zada, Azerbaijan University of Languages, Azerbaijan

The integration of Artificial Intelligence (AI) with decentralized computing paradigms continues to expand progressively. As emerging technologies evolve, this concept opens opportunities for building distributed intelligent systems that operate independently of any central authority. The primary objective of this research is to examine the architectural foundations of decentralized AI systems, investigate their major security challenges, and propose possible models aimed at preventing these problems. For this purpose, the interaction between federated learning, blockchain technologies, and edge computing is analyzed. Through these components, it becomes possible to achieve privacy-preserving collaboration, trustless coordination, and low-latency processing capabilities. However, they simultaneously introduce vulnerabilities such as model poisoning, Sybil attacks, and resource constraints. The obtained results lead to the conclusion that ensuring secure and scalable decentralized AI systems requires interdisciplinary approaches that combine artificial intelligence, cryptography, and distributed systems.

Blockchain, Federated Learning, Edge Computing, Sybil Attacks, Model Poisoning