A new way of detecting malicious code

A UC Berkeley computer scientist is developing mathematical algorithms based on semantics to help detect malicious code in computer viruses.

According to an article [1] about the research in Lab Notes, published by the university's College of Engineering, Sanjit Seshia and his colleagues are "using computational logic to detect the behavioral traits of viruses, even if their maliciousness is well-hidden by their creators."

The article says Seshia's work differs from current anti-virus software, which attempt to identify malware through a collection of rules and signatures. Such signatures become useless if virus writers change their code to thwart detection.

Seisha's method of detection searches the "meaning" of the code, says the article, as the algorithm can detect variations of malware. 

Read the full article here [1]. 

Linda Leung, lleung@nww.com [2]