Dynamic data encryption with polarized feedback

No Thumbnail Available
Authors
Shohrab, Saqlain
Issue Date
2023
Degree
Msc in Cybersecurity
Publisher
Dublin Business School
Rights
Items in eSource are protected by copyright. Previously published items are made available in accordance with the copyright policy of the publisher/copyright holder.
Abstract
In the ever-evolving domain of cryptography, achieving dynamic data encryption robustness remains paramount. This research paper delves into an innovative approach termed 'Dynamic Data Encryption with Polarized Feedback'. Leveraging neural network architectures, the study attempts to mimic the AES encryption and decryption process. After training the model for one million iterations using a singular AES key and cipher, the model achieved maximum accuracy when performing both encryption and decryption on its own. Intriguingly, when pitting the neural model against the traditional AES—encrypting with one and decrypting with the other—the success rate dropped significantly, achieving only 26 successful results out of 5,000 tests. This emphasizes the challenge in aligning neural encryption methods perfectly with conventional encryption techniques. The paper's code showcases an elegant interplay of PyTorch, the AES cipher, and a feedback mechanism that guides model retraining based on performance. The findings shed light on the complexities and potential pathways in neural cryptography.