Beacon Transcript – Google’ AI, part of the company’s Brain project, have shown the ability to come up with their own encryption methods which could mark a breakthrough, especially following the latest digital attacks.
As just last Friday saw another large distributed Denial of Service (DDoS) attack, everyone seems to be turning their eyes towards the need for a better cyber security system in both the usual digital devices and in the Internet of Things (IoT) domain.
With the number of cyber attack and attackers increasing just as fast as the technology, the best form of protection may come from the devices themselves.
As the AI (Artificial Intelligence) area continues to develop, Google recently revealed in a research paper that its AI’s have been able to develop their own encryption methods.
Google, the Internet giant, has established a separate division, the Google Brain, which seeks to research and advance the AI field, and has also established the Deep Learning project which targets neural networks.
Two of the company researchers, David Andersen and Martin Abadi, have reportedly assigned the AI’s the task to come up with their own encryption systems and techniques, and they succeeded.
Based on the neural network, which comprises massive databases and imitates the brain’s neural system, the AI’s were able to create their own encryptions, though they were somewhat basic when compared to the human-made ones.
The experiment fell on the Brain team’s three AI entities, named Alice, Eve, and Bob, all powered by the deep learning artificial neural networks.
The three were tasked to resolve a game which implicated a collaboration between the three as Alice had to send Bob an encrypted message, which he would then have to decode.
Eve would also have to intercept the message and decode it herself, without pre-determined data.
The initial message was a plain text which had to lose its meaning after the conversation so that only Bob could decrypt it, based on a set of pre-determined numbers, or keys, known by just the two of them.
Just as with humans, the first tries were quite unsuccessful as the message was as plain as day, but after almost 15,000 different attempts, the collaboration between Bob and Alice turned efficient as the AI’s worked out a decryption and encryption methods which could be read by both.
The final attempts also encumbered Eve, as only 8 of the 16 bits formed message could be intercepted and read.
The full practical implications of these encryption methods are as yet undetermined, as there are still a number of elements left to be understood as to how the method works.
However, this system could mark a first step towards a better security system as the digital age has started being the target of cyber attacks.
Image Source: Wikimedia