Post-Quantum TLS 1.3 Migration Framework with Adaptive Anomaly Detection for Encrypted Cloud Networks

Quantum computing poses a potential threat to classical cryptographic protocols, including Transport Layer Security (TLS) in cloud communications. In this paper, a Post-Quantum Cryptography-conscious Intrusion Detection System (PQ-IDS) is proposed to monitor encrypted communication using post-quantum TLS 1.3 securely. The framework incorporates CRYSTALS-Kyber-based key encapsulation into network traffic modeling. It presents an alternative data augmentation method that accounts for post-quantum characteristics, such as computation latency and communication overhead. To identify anomalies without decoding the payloads, a hybrid deep learning model comprising a Transformer, a Bidirectional LSTM, and attention mechanisms is used to learn temporal and spatial dependencies. Reinforcement learning is a dynamic approach to adaptive thresholding that optimizes the decision boundary in real time, enhancing detection robustness compared to fixed thresholds. Also, the explainable AI based on Integrated Gradients offers feature-level explanation. PQ-NF-CSE-CIC-IDS2018-v3 and the proposed PQ-CIC-PQC-OAV v1 data are experimentally evaluated, achieving a high F1-score of 90.9% and an ROC-AUC of 98%, low latency (0.036 ms/sample), and high throughput (22,649 samples/sec). The findings emphasize the usefulness of post-quantum awareness for intrusion detection from a computational-efficiency perspective and enable practical implementation in next-generation secure cloud settings.