New Software Is Teaching Computers to Detect Viruses
New Software Is Teaching Computers to Detect Viruses
July 17, 2014
Tech Page One — BGU researchers have begun developing and implementing a software that teaches computers to recognize foreign patterns and realize that they have a virus.
Much like the way a child tells his parents he is ill, this software can recognize when something is off internally, and can do so with a very low false positive rate.
Like an ill child however, even upon detection of malware, the threat is not eliminated. This technology does not replace anti-virus software, but works with it and makes it better.
The software is being developed by machine learning researchers, a branch of computer science which aims to teach computers how to think.
“The machine-learning approach applies learning algorithms that learn different patterns of benign and malicious files — not a specific signature — and therefore it is capable of detecting new malware,” says Prof. Lior Rokach, founder of the Machine Learning Research Lab, who is spearheading the project.
This means that this machine learning program can realize something in the programming is abnormal and can then alert you. Then, with the help of pre-existing anti-virus software, eliminate the threat.
According to Rokach and two colleagues at BGU, Prof. Yuval Elovici and Dr. Asaf Shabtai, who co-authored a study on the software, the false positive rate can be as low as five to ten percent.
The next step is to make the technology available on a mobile platform, a market which has seen little strides in the way of anti-virus software.
“Smartphones are a particularly valuable target for attackers because of the huge amount of personal information sitting on any given smartphone,” says Prof. Rokach.
“Additionally, a smartphone — unlike a desktop computer — can easily provide revenues to an attacker.”