Malware detection via data transformation monitoring

Abstract

Techniques and systems are described for detecting malware's bulk transformation of a user's data before the malware is able to complete the data transformation. Included are methods and systems for enabling malware detection by monitoring the file operations of a computer application or process for particular kinds of suspicious data transformation indicators. Indicators include primary indicators, such as file-type signature changes, notable changes in file data entropy, and out-of-range similarity measurements between the read and write versions of file data, as well as secondary indicators, such as a large number of file deletions and a large reduction in the number of file-types written versus read by a process over time. When indicators are triggered by a process, an adjustment to the process' malware score is made; in the event that the process' malware score reaches a malware detection threshold, the process is marked as malware and appropriate actions are taken.