Building better detection with privileged information

作者: Z Berkay Celik , Patrick D McDaniel , Rauf Izmailov , Nicolas Papernot , Ananthram Swami

DOI:

关键词:

摘要: Modern detection systems use sensor outputs available in the deployment environment to probabilistically identify attacks. These systems are trained on past or synthetic feature vectors to create a model of anomalous or normal behavior. Thereafter, run-time collected sensor outputs are compared to the model to identify attacks (or the lack of attack). While this approach to detection has been proven to be effective in many environments, it is limited to training on only features that can be reliably collected at test-time. Hence, they fail to leverage the often vast amount of ancillary information available from past forensic analysis and post-mortem data. In short, detection systems do not train (and thus do not learn from) features that are unavailable or too costly to collect at runtime. In this paper, we leverage recent advances in machine learning to integrate privileged information—features reliably available at training time, but not at run-time—into detection algorithms. We apply three different approaches to model training with privileged information: knowledge transfer, model influence, and distillation. We empirically validate their performance in a range of detection domains. Our evaluation shows that privileged information can increase detector precision and recall: we observe an average of 3.8% relative decrease in detection error for fast-flux bot detection over a system with no privileged information, 5.4% for malware traffic detection, 3.6% for malware classification, and 16.9% for user face authentication. We conclude by exploring the limitations and applications of different privileged information techniques in detection systems.

参考文章(0)