Neural Encryption Networks
I used a multiple autoencoders to effectively encode and encrypt text data. These provide encryption for user defined bits. As proof of concept 8 and 9 ...
deep-learning neural-cryptography machine-learning cyber-security tutorial
Objectives & Highlights

• Implemented a random hashing mechanism which hashes every utf encoding character to a random n bit character. • Weights and architecture of neural network are private keys. • Neural network ordering is public key. • Train multiple auto-enocoders that can encrypt this data. • Run them parallely to provide n bit fast encryption. • Hard to break encryption, as a word encrypted with letters of 8 bit 9 bit and 17 bit, which leads to exponential combinatorics problems.

Takeaways & Next Steps

• Can be used on top of RSA / AES to provide a great cyber-security framework. • Can even work on edge IoT devices as the encryption networks can be quantized.

Don't forget to add the tag @oke-aditya in your comments.

Machine Learning | Deep Learning | Data Science | Interested in developing new ideas.
Share this project