Address Resolution Protocol (ARP) spoofing attacks severely threaten Internet of Things (IoT) networks by allowing attackers to intercept, modify, or block communications. Traditional detection methods are insufficient due to high false positives and poor adaptability. This research proposes a multi-layered machine learning-based framework for intelligently detecting ARP spoofing in IoT networks. Our approach utilizes an ensemble of classifiers organized into multiple layers, each layer optimizing detection accuracy and reducing false alarms. Experimental evaluations demonstrate significant improvements in detection accuracy (up to 97.5\%), reduced false positive rates (less than 2\%), and faster detection time compared to existing methods. Our key contributions include introducing multi-layer ensemble classifiers specifically tuned for IoT networks, systematically addressing dataset imbalance problems, introducing a dynamic feedback mechanism for classifier retraining, and validating practical applicability through extensive simulations. This research enhances security management in IoT deployments, providing robust defenses against ARP spoofing attacks and improving reliability and trust in IoT environments.