A Dynamic Trust-Elimination Security Architecture for Sustainable Cyber-Threat Detection in Edge-Based IoT Systems
As the new applications like smart cities, medical systems, industrial automation, and critical infrastructures, along with the Internet of Things (IoT) technologies, have emerged, the need for real-time data processing at the edge has also grown. Although the edge-based IoT architectures minimize the latency and bandwidth consumption, they also present new security problems because of the distributed implementation, limited computational power, and lack of centralized control. The traditional, cloud-based, and black box artificial intelligence (AI) security system simply does not fit such a context without being transparent, responsible, and sustainable. Furthermore, in the real world, if IoT applications are designated to be safe, they cannot be realized by using black-box decision models without trust, compliance with regulations, and reliability through time. The paper proposes an edge-based IoT security framework to improve the cyber resilience and sustainability and ethical decision-making through the interpretable and responsible AI models. The proposed solution will be based on the use of explainable and light algorithms of machine learning at the network edge to monitor malicious actions and provide friendly explanations of the security actions to a human. It supports interpretability, fairness, accountability, and detection accuracy to detect threats and respond appropriately in dynamic environments of IoT, making the system accountable, fair, and sensible. The study with realistic IoT intrusion data sets proves that the proposed approach is competitive at detection and is more transparent and stable regarding various traffic conditions. The results show the importance of developing AI models that are interpretable and responsible for sustainable, credible, and resilient edge-based IoT security systems.