Machine Learning And AI Could Be The Future Of Cyber Security
IN-DEPTH: As the war against malware rages, smart software could be the key to getting ahead of hackers
In many ways this deep learning approach is more akin to moving from smart software to a proper AI.
Deep Insight is one cyber security company making waves with its deep learning approach to tackling the growing amount of malware, viruses and other malicious code that threaten everyone from enterprises to individuals.
Predictive threats
According to Guy Caspi, CEO of Deep Instinct, the company’s deep learning based real-time threat protection system is data agnostic and not beholden to searching for specific threat signatures or heuristic computing techniques found in traditional security software.
Nor does it require the same resources in terms of programming machine learning algorithms to tackle the near-continuous influx of new threats.
“[It] allows us to process any kind of data super-fast without any pre-processing like in machine learning, and it is in spired by the way our brain works,” he said. “The traditional methodologies cannot block this malware.”
Deep Instincts approach was to fire a huge amount of data at its deep learning system and allow it to identify existing cyber threats and the features that define them to automatically detect and prevent threat aimed at a network.
At first glance this does not differ too much from the machine learning techniques of Darktrace and Webroot. But given it is not trained to spot certain features that define malware, it has the ability to find new identifying aspects that could mean it figures out a threat before it becomes a danger.
Caspi noted that while the majority of anti-virus software on the market is generally capable at defending against known threats, Deep Instinct’s approach can curtail threats like zero-day flaws, which often have cyber security experts in a panic to plug.
“When it comes to very sophisticated vectors of attack……it becomes very challenging for these companies,” he said, noting the zero-day cat and mouse game between hackers and security experts is something Deep Instinct want to end.
“The idea of Deep Instinct was to change the paradigm of this cat and mouse game; to stop this and create something that match [hackers], and understand in advance any vectors of attack and create a very sophisticated system that will not need any process of signatures and heuristics.”
It aims to do this not by using existing threat intelligence but by having a team of hackers constantly trying to hack things and feeding that data into the deep learning neural net, which filters through the information to understand the vectors and methodologies hackers use to carry out their attacks and put in alerts or defences to mitigate them before they happen.
In essence, Deep Instinct does not so much detect threats as predict them and help ensure a company has the defences to mitigate even unknown threats.
Think of it as seeing someone for the first time, you might not know who they are but from the shape of their face and body you can tell whether they are a female or male; Deep Instinct essentially does this for cyber security.
The startup sized company claims to be the first firm to be using deep learning for cyber security, but if it can do what it claims you can be sure other will follow quickly.
