X-DeepID: An Explainable Hybrid Deep Learning Method for Enhancing IoT Security with Intrusion Detection

The Internet of Things (IoT) seamlessly integrates numerous devices with minimal human intervention, enabling effective communication between them, which has further enhanced the reliability across the diverse range of applications, including healthcare, intelligent agriculture, home security, industrial settings, and smart cities. However, the inherent nature of IoT infrastructure and its complex deployment components give rise to novel security challenges. Traditional security measures like encryption and access control prove insufficient in detecting attacks. Therefore, it becomes imperative to enhance the current security mechanisms in order to establish a safeguarded IoT environment. The progression of deep learning techniques introduces embedded intelligence to the IoT domain, addressing various security concerns. Nonetheless, these machine learning models tends to yield elevated false-positive rates, due to which comprehending the reasons behind its forecasts can be complex, even for experienced experts. The capacity to grasp the rationale behind an Intrusion Detection System's (IDS) choice to block a specific packet is crucial for cyber security professionals. This comprehension empowers them to verify the system's efficacy and to engineer more robust and cyber-resilient systems. In this work, we put forth a framework for intrusion detection system which is based on explainable hybrid deep learning model. This framework aims to enhance the clarity and robustness of Deep Learning-based IDS within IoT networks. The experimental findings have showcased the exceptional achievement of the proposed work, achieving an impressive accuracy of 99.25% and an outstanding F1 score of 98.91%. These results vividly demonstrate the framework's capacity to safeguard IoT networks effectively against sophisticated cyber attacks.