AI & Machine learning

 

We came across an interesting headline that suggested AI could be the answer to the rise in hacking, particularly with the eagerly anticipated Internet of Things.

 

Darpa (Defence Advanced Research Project Agency) has been bringing engineers together for what they called ‘Grand Challenges’. Their next challenge aims to develop software smart enough to seek and seal areas that are susceptible to hackers in other software. It aims to seal the holes before the hacker is aware of the vulnerability. This could particularly help with zero-day threats and as mentioned earlier it would benefit IoT as the world is increasingly being  connected by lots of small, smart-net devices.  

 

Currently, the process for plugging vulnerabilities is manual, reactive and slow. If the IoT is to grow at the rate that is expected then the current process for fixing exposures will struggle to keep devices protected.

 

Automated smart defences are already in use, in the form of anti-virus software. Traditional anti-virus software was woeful at handling malware it hadn’t seen before. Trojans, worms and other type of viruses are made in the millions everyday, so it was a necessity that detection became automatic. Some of the ways of catching these new viruses is to use advanced software to generalise code it did know to spot the code it didn’t.

 

In truth, this is machine learning as opposed to actual AI. It is a step towards AI but there is still the need for human invention. It will get very interesting when malware starts machine learning.
Google last month was reported to have used machine learning and cut data centre energy by 15%. This reduction was accomplished through a combination of more accurately predicting incoming computational loads and then quickly adjusting the cooling load. Google doesn’t disclose how much electricity the data centres use, but as a company, it has said it is responsible for 0.01% of the global electricity use. Following in the footsteps of Google, we’re also currently in the process of using software tools to track energy usage and leakage to reduce energy consumption in our data centre and will share the results shortly here.