Nvidia Invests In Deep Instinct, A Deep Learning-Based Cybersecurity Startup

Nvidia announced that it has invested in Deep Instinct, an Insrael-based startup that uses deep learning technology to detect and prevent the “most advanced cyberattacks.”

AI For Cybersecurity

Deep learning is a promising machine learning technology  used across industries. Many companies are now using it to train neural network models that achieve high accuracy without too much help from programmers in specifying what the rules should be for a given task.

Academics researching AI that can stop cyberattacks have previously shown an 85% success rate in detecting cyberattacks. Although that number may look impressive, it means that one out of seven attempts to attack your systems will be successful.

It’s also not much better than what you’d obtain from an antivirus program that doesn’t use machine learning to stop attacks. In fact, most of the top antivirus programs seem to be doing much better than that.

The latest report from AV-Comparatives on how the antivirus programs fare against real threats, they all score from 92% to 100%. However, these are automated tests, and they may not account for the full range of attacks that a system may see, especially regarding advanced cyberattacks from nation-state actors.

Deep Instinct’s Solution

Deep Instinct, which claims to be the "first company to apply deep learning to cybersecurity," has also stated that its deep-learning solution can defeat 99% of “cyberattacks." That includes both known malware as well as “first-seen” attacks against a system.

Nvidia vice president of business development Jeff Herbst said:

Deep Instinct is an emerging leader in applying GPU-powered AI through deep learning to address cybersecurity, a field ripe for disruption as enterprise customers migrate away from traditional solutions. We’re excited to work together with Deep Instinct to advance this important field.

Due to its deep learning based technology, Deep Instinct’s solution can run locally on any type of device, and it doesn’t require a connection for cloud analysis. This should keep the data on the devices private, as the file analysis is not sent over to the company’s servers.

The solution is also lightweight, because it doesn’t have to check against hundreds of thousands of malware signatures in real-time. All it needs is the neural network--which was trained against hundreds of millions of malicious and benign files--and a powerful enough GPU to run the model.

Deep Instinct was recently named a “Technology Pioneer” by The World Economic Forum, and the “Most Disruptive Startup” at NVIDIA’s 2017 Inception Awards.

Lucian Armasu
Lucian Armasu is a Contributing Writer for Tom's Hardware US. He covers software news and the issues surrounding privacy and security.