Google is turning to the use of machine learning in a bid to limit the impact of malware on Android devices. The information comes by way of a report out of The Register, which in turn draws on information provided by Android's security head Adrian Ludwig, who was explaining the advancements made through the use of machine learning during a talk at the Structure Security conference in San Francisco.
According to those details, the difference with the use of artificial intelligence is that it looks to identify threats by their behavior and through the use of information drawn from devices in the wild. The notion put forward here is that as the underlying technology progresses, it gets better at identifying malware traits and can use that continuously-increasing knowledge to identify malware apps through the code’s actions, rather than their actual code. Which according to Ludwig, is already starting to pay dividends.
Ludwig points out that at the start of 2017 the software in question was only able to identify around 5-percent of the malware that it was purposely exposed to. With 2017 still far from over that number is now said to have increased to 55-percent. Likewise, Ludwig goes on to explain roughly over the same time the percentage of Android's user base infected by malware had dropped from 0.6-percent (earlier in 2017) down to a current figure closer to 0.25-percent. Suggesting the longer the software is in place and being used, the quicker it is getting at understanding the malware and identifying it. Although not explicitly suggested, the wider speculation here is that the software may be able to eventually identify malware in advance of seeing its behavior by identifying common traits between the newly arriving apps and those that it has previously identified as ‘bad’ apps. Of course, this is only one of the ways in which Google has turned to machine learning as the company has been applying the use of artificial intelligence to many services of late, with a bid to improving them all to some capacity. So while it might not be too surprisingly to learn that Android security is another one of those services now making use of machine learning, early indications suggest that it might be one of the services which could really benefit from the technology, and quickly.