An AI-Driven Cybersecurity Framework for IoT: Integrating LSTM-Based Anomaly Detection, Reinforcement Learning, and Post-Quantum Encryption

Abstract

The number Internet of Things (IoT) devices are growing rapidly. Due to this many new IoT cybersecurity challenges are being introduced. The challenges such as limited computational power on the IoT devices leading to them being continuously exposed to changing network threats making IoT Security difficult. Existing security approaches use techniques like deep learning for intrusion detection, reinforcement learning for adaptive response, data integrity mechanisms, and post-quantum cryptography. However, these techniques are often studied separately. But integrating these techniques still faces some challenges such as higher resource consumption, generalization, adaptive response and practical deployment. This paper proposes an AI driven framework that integrates LSTM based anomaly detection, Q-learning based adaptive response, data integrity verification, and post-quantum cryptographic protection. The framework that has been proposed is designed to address some of the challenges such as to detect abnormal behaviour in the network. The framework will also assess the threat level and select the appropriate response and provide additional security to the IoT devices. This framework will be evaluated using datasets such as CICIoT2023 and Edge-IIoTset, with the main focus on computational overhead, performance of detection system, cross-dataset generalisation and adaptive response. This security system approach aims to provide improved resilience and security for IoT environments with consideration of limited resources.