Using Temporal Attributes to Detect Malware

Techniques for classifying unknown files taking into account temporal proximity between unknown files and files with known classifications are disclosed. In response to a classification request for a target file, client systems hosting (or hosted) instances of the target file are identified. For each system, files created around the time the target file was created on the system are identified. Within the identified files, files with known classifications are identified, and a score is determined for each such file to measure temporal proximity between the creation of the file and the creation of the target file. Local temporal proximity scores aggregate the scores for the client system. Global temporal proximity scores measures an aspect of the local temporal proximity scores for all identified client systems. The global temporal proximity scores are fed into a classifier to determine a classification, which is returned in response to the classification request.