Accessing Encrypted Information with Deep Learning
Elina Tlachenska, Kiril Ivanov, Maria Nenova, Rumen Doynov, Zlatka Valkova-Jarvis · 2025
With the application of artificial intelligence in various sectors, paramount has become the security concerns. Artificial intelligence systems often lack robust security measures, exposing them to data breaches and privacy violations. Cryptography has rendered security by ensuring authentication data confidentiality. Cryptographic techniques are becoming necessary for sensitive data and algorithms as the application of AI distributes over industries including finance, cybersecurity, and healthcare. The goal of this paper is to present deep learning and cryptography. Their basic attributes and functions through which they are used. To consider methods for accessing encrypted information with deep neural networks and finding (guessing) a plaintext. The effectiveness of an approach with simple classical ciphers will be shown by presenting a substitution attack on a cipher by a neural network.