Automated Cryptographic Weakness Discovery: A Reinforcement Learning Approach for Adaptive Cryptanalysis
B Kiran Kumar, Ajay Kumar, M. Rekha Sundari, Dondapati Tejaswi, Dogga Ashwani, Saurabh Bilgaiyan · 2025
Today's cryptographic algorithms maintain security through defense against multiple attack approaches, but advancing encryption protocols leads to advanced attack methods. A breakthrough cryptanalysis system based on reinforcement learning technology is presented in this research to detect weaknesses in cryptographic systems automatically. The proposed reinforcement learning system improves its attack methods through learning from previous tests and continuously becomes more proficient. Our team evaluates this methodology against blockchain ciphers like AES while testing public key usage like RSA. The model can discover sophisticated mistakes that resist detection with traditional cryptanalysis methods. The adaptable and expandable framework provides the foundation for future cryptography exploration and automated attack creation methods.