Abstract

We propose an encryption communication system based on the adversarial networks, called Adversar; Network Encryption System (ANES), which contains Sender, Receiver and various Attackers. Using a general loss function, we design three types of ANES with different Attackers, which simulate different password attacks on communications. Through the one-to-one adversarial training between Sender and these all Attackers, ANES can frequently master One Time Pad (OTP) algorithms without human cognition. We find that the more and stronger the opponents ANES encountered, the higher the probability of learning OTP. We also present a more aggressive attack network based on Chosen Ciphertext Attack (CCA) as the antagonist of Sender and Receiver, and finally the system learns OTP with a considerable probability. Experimental results demonstrate that our encryption strategy is comparable to other related works.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call