AI Powered Crypt-Analysis for Identification of Encryption Algorithm

In modern world, cryptographic algorithms are critical for securing sensitive data, ensuring confidentiality, and protecting against cyber threats. With the increasing number of encryption algorithms and modes, identifying which algorithm was used in a ciphertext becomes a difficult job. To tackle this we present a machine learning-based solution that automates the identification of cryptographic algorithms using the Random Forest classifier. The system processes a large dataset of “encrypted-data”, consisting of 39 different algorithms and modes, by first extracting key features such as byte frequency and entropy from the ciphertexts. These features are used to create a 257 -length cipher-descriptor for classification. The dataset is split into 19 subsets, each containing data from two algorithms, and 19 different Random Forest models are trained. During testing, a sequential model loading approach is employed to manage computational resources. The best-performing model, in terms of confidence score, is used to predict the encryption algorithm for a given input ciphertext. Finally, the system is integrated with a Gradio web interface, allowing for real-time, user-friendly interaction. Our approach provides a scalable and accurate tool for cryptographic analysis, significantly improving upon traditional methods.