Paper at Usenix Security Symposium

Our work on detecting concept drift in malware classification models titled “Transcend: Detecting Concept Drift in Malware Classification Models” has been accepted at Usenix Security Symposium 2017. Concept drift in malware classification is a serious issue whereby models trained on older malware are unable to classify newer malware with confidence. This work aims at providing hints on identifying such antiquated classification models which are harbingers of poor classification decisions.