Neural Network-Based Double Encryption for Jpeg2000 Images

• Main Author: Memon, Qurban Ali
• Format: Article
• Language: English
• Published: Universiti Utara Malaysia Press 2017
• Subjects: QA75 Electronic computers. Computer science
• Online Access: https://repo.uum.edu.my/id/eprint/29511/1/JICT%2016%2001%202017%20137-155.pdf; https://repo.uum.edu.my/id/eprint/29511/; https://e-journal.uum.edu.my/index.php/jict/article/view/8226

The JPEG2000 is the more efficient next generation coding standard than the current JPEG standard. It can code files witless visual loss, and the file format is less likely to be affected by system file or bit errors. On the encryption side, the current 128-bit image encryption schemes are reported to be vulnerable to brute force. So there is a need for stronger schemes that not only utilize the efficient coding structure of the JPEG2000, but also apply stronger encryption with better key management. This research investigated a two-layer 256-bit encryption technique proposed for the JPEG2000 compatible images. In the first step, the technique used a multilayer neural network with a 128-bit key to generate single layer encrypted sequences. The second step used a cellular neural network with a different 128-bit key to finally generate a two-layer encrypted image. The projected advantages were compatible with the JPEG2000, 256-bit long key, managing each 128-bit key at separate physical locations, and flexible to opt for a single or a two-layer encryption. In order to test the proposed encryption technique for robustness, randomness tests on random sequences, correlation and histogram tests on encrypted images were conducted. The results show that random sequences pass the NIST statistical tests and the 0/1 balancedness test; the bit sequences are decorrelated, and the histogram of the resulting encrypted images is fairly uniform with the statistical properties of those of the white noise.