Saturday, July 8, 2017

AI SPECIAL.... Tech vs tech: How AI is taking on ransomware

Tech vs tech: How AI is taking on ransomware


Experts Use Machine Learning Tools To Identify Malicious Software & Nip Attacks In The Bud

Twice in the space of six weeks, the world has suffered major attacks of ransomware -malicious software that locks up photos and other files stored on your computer, then demands money to release them.
It's clear that the world needs better defences, and fortunately those are starting to emerge, if slowly . When they arrive, we may have artificial intelligence to thank.
Despite the risks of ransomware attacks, many people just aren't good at keeping up with security software updates. Both recent attacks walloped those who failed to install a Windows update. Watchdog security software has its problems, too.With this week's attack, only two of about 60 security services tested caught it at first, according to researchers. “A lot of normal applications, especially on Windows, behave like malware, and it's hard to tell them apart,“ said Ryan Kalember, an expert at the US security vendor Proofpoint.
In the early days, identifying malicious programs involved matching their code against a database of known malware. But this technique was only as good as the database; new variants could easily slip through.
So security companies started characterising malware by its behaviour. In the case of ransomware, the software could look for repeated attempts to lock files by encrypting them. But that can flag ordinary computer behaviour such as file compression. Newer techniques involve looking for combinations of behaviours. For instance, a program that starts encrypting files without showing a progress bar could be flagged for surreptitious activity , said Fabian Wosar, chief technology officer at the New Zealand security company Emsisoft.But that also risks identifying harmful software after some files have been locked up.
An even better approach identifies malware using observable characteristics usually associated with malicious intent -for instance, by quarantining a program disguised with a PDF icon.
This sort of malware pro filing wouldn't rely on exact code matches, so it couldn't be easily evaded. And such checks could be made well before dangerous programs start running.
Still, two or three characteristics might not properly distinguish malware from legitimate software. But how about dozens? Or even thousands? For that, researchers turn to machine learning, a form of artificial intelligence. The security system analyses samples of good and bad software and figures out what combination of factors is likely to be present in malware. As it encounters new software, the system calcu lates the probability that it's malware, and rejects those that score above a certain threshold. When something gets through, it's a matter of tweaking the calculations or adjusting the threshold.
On the flip side, malware writers can obtain these tools and tweak their code to see if they can evade detection.Some websites already offer to test software against leading security systems. Still, security firms employing machine learning have claimed success in blocking most malware, not just ransomware. SentinelOne even offers a $1million guarantee against ransomware; it hasn't had to pay it yet.

AP

No comments: