The rapid expansion of the Internet of Things (IoT) has introduced a new era of interconnected intelligence, enabling seamless automation across homes, industries, and cities. However, this massive connectivity also exposes IoT ecosystems to complex and evolving cyber threats that traditional static defense mechanisms are unable to counter effectively. To address these vulnerabilities, this research proposes a novel cyber defense framework built upon self-learning artificial intelligence (AI) agents capable of autonomously detecting, adapting to, and mitigating malicious activities within dynamic IoT environments. The proposed system integrates adaptive learning techniques that enable agents to evolve their decision-making capabilities based on environmental feedback and observed threat behaviors. Through continuous interaction and shared learning, these agents collectively enhance network resilience by predicting potential attack patterns before they materialize. Experimental evaluations conducted within simulated IoT networks demonstrate significant improvements in detection accuracy, response efficiency, and adaptability when compared to conventional rule-based systems. The study underscores the transformative potential of autonomous AI-driven defense mechanisms in ensuring secure and resilient IoT infrastructures. The outcomes contribute to the growing discourse on intelligent cybersecurity by highlighting how self-learning models can redefine proactive defense strategies in the age of pervasive digital interconnectivity.

Keywords: Self-Learning AI Agents, Adaptive Cyber Defense, Internet of Things (IoT), Reinforcement Learning, Multi-Agent Systems, Intrusion Detection, Cybersecurity Automation