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

In an increasingly digital world, cybersecurity threats have grown in frequency, complexity, and scale, challenging the adequacy of traditional rule-based defense systems. This study proposes a unified AI-driven cybersecurity framework to detect anomalies, verify data integrity, automate incident response, and ensure long-term cryptographic resilience. The framework integrates four core components: Long Short-Term Memory (LSTM) networks for temporal anomaly detection, homomorphic hashing using SHA-256 with hidden salts for real-time data integrity verification, Q-learning-based reinforcement learning for automated threat response, and lattice-based encryption grounded in the Learning With Errors (LWE) problem to safeguard against quantum-era attacks. The system was evaluated in a simulated IoT network environment, where it demonstrated high accuracy in identifying anomalies, effectively distinguishing between original and tampered data, and adaptively responding to different levels of cyber threats. The integration of these components allows the framework to operate autonomously and contextually, improving scalability and responsiveness in resource-constrained digital infrastructures. This study concludes that the proposed framework addresses key limitations of existing methods by offering a scalable, adaptive, and future-proof cybersecurity solution. The results support its potential for deployment in real-world settings such as smart cities, healthcare systems, and critical infrastructure, with future work aimed at improving real-time adaptability and validating performance in live, heterogeneous environments.