Ransomware Variant Eludes Detection Machine Learning Algorithms | SC Media

Ransomware Variant Eludes Detection Machine Learning Algorithms

March 29, 2017
By Marcos Colon

A variant of the Cerber ransomware has made itself much harder to detect by adopting a new technique that thwarts a popular detection mechanism.

Researchers at security firm Trend Micro have discovered that the variant can evade detection by machine learning solutions, according to a recent alert issued by the company.

Similar to other ransomware attacks, this variant of Cerber is distributed via email. But the malware differs from others in that it separates its different stages into multiple files to evade how machine learning solutions operate.

“The industry has created features to proactively detect malicious files based on features instead of signatures,” wrote Gilbert Sison, team manager at Trend Micro. “The new packaging and loading mechanism employed by Cerber can cause problems for static machine learning approaches-i.e, methods that analyze a file without any execution or emulation.” 

prestitial ad