Google has clearly indicated that AI can create its cryptographic standard – which is indeed a critical step towards automating encryption. In a paper released by the Google team, it indicates that the three neutral networks developed – Alice, Eve, and Bob – were able to carry out the individual tasks that were assigned to them.
For instance, Eve (the hacker) was able to crack the message sent to Bob for decrypting by Alice, the message sender. The hacker was asked to crack the message without using any key. Here, all the three networks were asked to make up the message as they went along. Google team was able to demonstrate that it was possible to train Alice and Bob to defeat the strongest version of Eve.
15, 0000 simulations were carried out, and Bob together with Alice were able to decrypt and send secure messages, but Eve remained unsuccessful in trying to decrypt the messages. In another exercise – asymmetric encryption – things turned out to be a bit difficult. In this case, the outcomes were different and harder to analyse. It was only when Eve got trained that it became easier for him to decrypt the messages sent between Alice and Bob. However, Alice and Bob could communicate secretly.
The exercises were simple; they did not include cryptographic algorithms to come up with complex systems, according to the academic paper presented. This is particularly true since the networks used a set of secrecy specification backed up with training objectives. While the system remains simple, the research indicate that the method appears dynamic with the capability of gaining strength as hackers try to crack it.
The networks are set to drastically increase security; it will be interesting to observe the level of adoption among businesses – but its future looks clear and quite promising.