Exact constrained-training neural networks for confidential 8-bit arithmetic primitives in code obfuscation

Key Points

• Enhanced data confidentiality achieved through the use of constrained neural networks in arithmetic primitives, enabling secure code obfuscation.
• Performance metrics indicate significant improvement in the efficiency of 8-bit arithmetic operations, with lower error rates compared to traditional methods.
• Constrained training was utilized to optimize neural networks specifically for code obfuscation applications in this study.
• The study highlights the potential for future developments in secure computing practices, emphasizing the need for advanced neural network algorithms.