Despite state-of-the-art antivirus software available in the market, malware and virus-related security breaches cost more than $500 billion each year worldwide.
Modern-day antivirus software can easily protect an average consumer from all sorts of viruses. Most of the antivirus software virus databases are often automatically updated (the program allows you to tweak the settings). The problem is, sometimes virus attacks happen when your virus database is still to be updated. After the attack although your antivirus can help you get rid of the virus sometimes by the time your database is updated, the virus has done the damage. How do you stop that?
By using an antivirus software that does not have to update the virus database in order to protect you from new viruses. Can this be done?
Researchers are using deep learning algorithm to detect new malicious code naturally, without database updates. It is like recognising a face in an image or a photograph. If such a complex algorithm that recognises faces can be developed, then why not have an algorithm that recognises viruses?
According to an Israeli start-up Deep Instinct, such an algorithm can be developed. The company is developing a deep-learning antivirus that can spot new viruses and malware with 20% higher accuracy than even the best contemporary virus protection software. According to an experiment done by the start-up, the software could detect 95% of new viruses without having to be updated. Although the antivirus software is still in the early stages, it can be a big game changer. It can spell nd end for prospective virus developers because their virus will be caught before it can cause damage.
According to this MIT technology review, Deep Learning antivirus software mimics the brain in order to catch malware. Artificial neural networks can be trained to recognise characteristics of malicious code by going through millions of examples of malware and non-malware files. This approach is called deep learning.