Harnessing AI and Post-Quantum Cryptography for Cybersecurity in the Quantum Era

The talk opens with a state-linked group that quietly moves sensitive data out through web traffic. Quantum computers would eventually break RSA and ECC through Shor's algorithm and weaken symmetric keys through Grover's, though AES-256 remains considered post-quantum safe. Meanwhile, harvest-now, decrypt-later campaigns already collect long-lived data such as intellectual property and government secrets for future decryption.

Conventional tools such as IDS, IPS, DLP and next-generation firewalls catch known threats but miss zero-days and stealthy intruders. AI can add behavioral analytics, adaptive firewall tuning and predictive threat modeling. The speaker compared isolation forests, autoencoders and variational autoencoders on a network intrusion dataset with about 18 features, and autoencoders performed best on accuracy, precision and recall. Precision keeps false alarms manageable, while high recall is key for catching slow, silent exfiltration.

The transition plan starts with a cryptographic inventory covering algorithms, key sizes, rotation practices and how long protected data must stay confidential. Teams should build crypto agility through modular libraries, consider hybrid schemes during migration, and move toward the NIST-standardized ML-KEM, ML-DSA and SLH-DSA algorithms while following CNSA and NIST guidance. The closing message was that these changes are continuous rather than one-time, so planning should begin now.