Adaptive Latent Representation Learning with Consensus Decision Intelligence for Trusted IoT Device Identity Verification

Abstract

The continuous growth of the Internet of Things (IoT) has led to the deployment of a vast ecosystem of interconnected devices across domains such as healthcare, smart cities, industrial automation, and intelligent infrastructure. While this connectivity offers significant operational benefits, it also introduces serious cybersecurity risks, including unauthorized access, data tampering, malicious attacks, and network service interruptions. Conventional security solutions, such as Access Control Lists (ACLs), signature-based detection techniques, and ruledriven mechanisms, are increasingly inadequate because of their fixed configurations, dependence on continuous manual updates, and inability to effectively recognize newly emerging threats. To overcome these limitations, this research proposes an intelligent IoT security framework based on a Deep Autoencoder (DAE) that performs both device classification and data integrity verification. The DAE learns compact and informative feature representations from IoT network traffic, enabling accurate differentiation between legitimate and malicious device activities. By utilizing labeled datasets during training, the model captures complex communication patterns and detects anomalous behaviour with high precision. In addition, an integrated authentication module validates the authenticity and integrity of device communications before allowing access to the network, thereby strengthening overall system trust and security. Experimental evaluation demonstrates that the proposed framework achieves an accuracy of 97.8%, surpassing conventional machine learning techniques such as K-Nearest Neighbors (KNN) and Logistic Regression Classifier (LRC). Furthermore, a hybrid model referred to as DAE-BFL (BFL), which combines the Deep Autoencoder with Random Forest Classifier (RFC) and Logistic Regression Classifier (LRC), further improves classification reliability and robustness. The overall framework encompasses data preprocessing, model training, validation, and real-time performance assessment, providing a scalable, adaptive, and efficient solution for securing next-generation IoT networks.