Academic Publications
Threat of Offensive AI To Organizations
Computers and Security, 2022
TL;DR: There are 32 ways an adversary can use AI to automate their tools, tactics and techniques so as to attain their objectives
Yisroel Mirsky, Ambra Demontis, Jaidip Kotak, Ram Shankar Siva Kumar, Deng Gelei, Liu Yang, Xiangyu Zhang, Maura Pintor, Wenke Lee, Yuval Elovici, Battista BiggioICML Workshop on Adversarial Machine Learning, 2021
TL;DR: Adversarial ML researchers believe attacks are “bad” and defenses are “good” without the context. They believe their work is neutral when it is not.
Kendra Albert, Maggie Delano, Bogdan Kulynych, Ram Shankar Siva KumarEthical Testing in the Real World: Evaluating Physical Testing of Adversarial Machine Learning
NeurIPS Workshop on Dataset Curation, 2020
NeurIPS Workshop on Navigating the Broader Impacts of AI Research, 2020
TL;DR: Papers on physical adversarial attacks like adversarial cap, adversarial tshirt, adversarial glassess characterize themselves as "real world." Despite this framing, however, we found the physical or real-world testing conducted was minimal, provided few details about testing subjects and was often conducted as an afterthought or demonstration.
Kendra Albert, Maggie Delano, Jonathon Penney, Afsaneh Rigot, Ram Shankar Siva KumarReflecting on Paradise Lost via Reinforcement Learning and Resistance AI Literature
NeurIPS Workshop on Resistance AI, 2020
TL;DR: Anachronistic parallels between John Milton’s 350 year old poem and Reinforcement Learning
Ram Shankar Siva KumarLegal Risks of Adversarial Machine Learning Research
ICML Workshop on Law and Machine Learning, 2020
TL;DR: We use the flagship federal anti-hacking law, the Computer Fraud And Abuse Act (CFAA), to analyze attacks on AI systems.
Ram Shankar Siva Kumar, Jonathon Penney, Bruce Schneier, Kendra AlbertAdversarial machine learning-industry perspectives
IEEE Security and Privacy Workshop, 2020
TL;DR: 25 out of 28 organizations we surveyed do not know about attacks on AI systems
Ram Shankar Siva Kumar, Magnus Nyström, John Lambert, Andrew Marshall, Mario Goertzel, Andi Comissoneru, Matt Swann, Sharon XiaPolitics of adversarial machine learning
ICLR Workshop on Trustworthy ML, 2020
TL;DR: Attacks on AI systems have political dimensions (in the Langdon Winner sense; not republican/democractic). That is, attacks on AI systems enable or foreclose certain options for both the subjects of the machine learning systems and for those who deploy them, creating risks for civil liberties and human rights
Kendra Albert, Jon Penney, Bruce Schneier, Ram Shankar Siva KumarHardening quantum machine learning against adversaries
New Journal of Physics, 2018
TL;DR: Quantum machine learning systems are also fallible to adversarial examples and can be disrupted by an adversary
Nathan Wiebe, Ram Shankar Siva KumarLaw and Adversarial Machine Learning
NeurIPS Workshop on Security in Machine Learning, 2018
TL;DR: We explore how some aspects of computer crime, copyright, and tort law interface with adversarial ML
Ram Shankar Siva Kumar, David R O'Brien, Kendra Albert, Salome VilojenPractical machine learning for cloud intrusion detection: challenges and the way forward
ACM Workshop on Artificial Intelligence and Security, 2017
TL;DR: Operationalizing machine learning based security detections is extremely challenging, especially in a continuously evolving cloud environment.
Ram Shankar, Andrew Wicker, Matt Swann