Crypto-ransomware Detection through Quantitative API-based Behavioral Profiling

Wenjia Song, Sanjula Karanam, Danfeng (Daphne) Yao

Abstract

Crypto-ransomware has caused an unprecedented scope of impact in recent years with an evolving level of sophistication. We are in urgent need to pinpoint the security gap and improve the effectiveness of defenses by identifying new detection approaches. In this paper, we quantitatively characterized the runtime behaviors of 54 ransomware samples from 35 distinct families, with a focus on the core encryption and file access behaviors. Based on our characterization results on dynamic API behaviors of ransomware, we present a new API profiling-based detection mechanism. Our method consists of two operations, namely consistency analysis and API-contrast-based refinement. We evaluate our approach against a set of real-world ransomware as well as benign samples. The results show that our method effectively detects all ransomware executions in the consistency analysis and further reduces the false positive case in API-contrast refinement. We also conduct in-depth case studies to investigate the most informative APIs for detection with context.