This thesis aimed to explore Sybil attack detection in Machine-to-Machine payment systems through a layered architectural approach using Multi-Agent Systems and Blockchain. While the results did not show outstanding performance, they revealed important insights into the dynamics of detection strategies, showing that combining different detection approaches can reduce false positives in environments with multiple types of attackers. The experimental outcome serves as a proof-of-concept for how this architectural synergy can be used to detect and punish Sybil attackers.