Self-Supervised Deep Learning for Secure and Efficient IoT Device Authentication

The rapid expansion of the Internet of Things (IoT) has introduced significant security challenges, particularly in device authentication and identity verification. Traditional authentication mechanisms, such as passwordbased and certificate-based approaches, are often vulnerable to attacks and unsuitable for resource-constrained IoT environments. This paper proposes a self-supervised deep learning framework for secure and efficient IoT device authentication, leveraging contrastive learning and anomaly detection techniques. The proposed model learns devicespecific representations without requiring labeled data, enabling robust authentication even in dynamic IoT ecosystems. By employing transformer-based encoders and contrastive loss functions, the framework effectively distinguishes legitimate devices from adversarial entities. Furthermore, an edge-computing deployment strategy ensures low-latency authentication, making it scalable for large-scale IoT networks. Experimental results demonstrate that the proposed model outperforms conventional supervised approaches in accuracy, robustness, and computational efficiency. This work highlights the potential of self-supervised deep learning in enhancing IoT security while minimizing data annotation efforts, paving the way for future autonomous authentication systems.