UC Davis

Malware behavior detectors observe the behavior of suspected malware by emulating its execution or executing it in a sandbox or other restrictive, instrumented environment. This assumes that the process, or process family, being monitored will exhibit the targeted behavior if it contains malware. We describe a technique for evading such detection by distributing the malware over multiple processes. We then present a method for countering this technique, and present results of tests that validate our claims.