2021/11/12 by Daniel Steinberg, Steinberg, Daniel, Paul Munro +1
Computer Science · #Advanced Malware Detection Techniques #Adversarial Robustness in Machine Learning #Anomaly Detection Techniques and Applications #Cryptography and Security (cs.CR) #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (cs.LG)
paper · pdf · doi:10.48550/arxiv.2111.07035
openalex publication_date 2021/11/12 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Deep learning models have been used for a wide variety of tasks. They are\nprevalent in computer vision, natural language processing, speech recognition,\nand other areas. While these models have worked well under many scenarios, it\nhas been shown that they are vulnerable to adversarial attacks. This has led to\na proliferation of research into ways that such attacks could be identified\nand/or defended against. Our goal is to explore the contribution that can be\nattributed to using multiple underlying models for the purpose of adversarial\ninstance detection. Our paper describes two approaches that incorporate\nrepresentations from multiple models for detecting adversarial examples. We\ndevise controlled experiments for measuring the detection impact of\nincrementally utilizing additional models. For many of the scenarios we\nconsider, the results show that performance increases with the number of\nunderlying models used for extracting representations.\n