A Kings Ransom for Encryption: Ransomware Classification using Augmented One-Shot Learning and Bayesian Approximation

摘要: Newly emerging variants of ransomware pose an ever-growing threat to computer systems governing every aspect modern life through the handling and analysis big data. While various recent security-based approaches have focused on detecting classifying at network or system level, easy-to-use post-infection classification for lay user has not been attempted before. In this paper, we investigate possibility a is infected with simply based screenshot splash screen ransom note captured using consumer camera commonly found in any mobile device. To train evaluate our system, create sample dataset screens 50 well-known variants. dataset, only single training image available per ransomware. Instead creating large screenshots, simulate capture conditions via carefully designed data augmentation techniques, enabling simple efficient one-shot learning. Moreover, model uncertainty obtained Bayesian approximation, ensure special input cases such as unrelated non-ransomware images previously-unseen are correctly identified mis-classified. Extensive experimental evaluation demonstrates efficacy work, accuracy levels up 93.6% classification.