Search Results for « AI/ML Security »

This paper focuses on supporting AI/ML Security Workers -- professionals involved in the development and deployment of secure AI-enabled software systems. It presents AI/ML Adversarial Techniques, Tools, and Common Knowledge (AI/ML ATT&CK) framework to enable AI/ML Security Workers intuitively to explore offensive and defensive tactics.

In this case study, we describe the design and assembly of a cyber security testbed at Oak Ridge National Laboratory in Oak Ridge, TN, USA. The range is designed to provide agile reconfigurations to facilitate a wide variety of experiments for evaluations of cyber security tools -- particularly those involving AI/ML. In particular, the testbed provides realistic test environments while permitting control and programmatic observations/data collection during the experiments. We have designed in the ability to repeat the evaluations, so additional tools can be evaluated and compared at a later time. The system is one that can be scaled up or down for experiment sizes. At the time of the conference we will have completed two full-scale, national, government challenges on this range. These challenges are evaluating the performance and operating costs for AI/ML-based cyber security tools for application into large, government-sized networks. These evaluations will be described as examples providing motivation and context for various design decisions and adaptations we have made. The first challenge measured end-point security tools against 100K file samples (benignware and malware) chosen across a range of file types. The second is an evaluation of network intrusion detection systems efficacy in identifying multi-step adversarial campaigns -- involving reconnaissance, penetration and exploitations, lateral movement, etc. -- with varying levels of covertness in a high-volume business network. The scale of each of these challenges requires automation systems to repeat, or simultaneously mirror identical the experiments for each ML tool under test. Providing an array of easy-to-difficult malicious activity for sussing out the true abilities of the AI/ML tools has been a particularly interesting and challenging aspect of designing and executing these challenge events.

When `cyber' is used as a prefix, attention is typically drawn to the technological and spectacular aspects of war and conflict -- and, by extension, security. We offer a different approach to engaging with and understanding security in such contexts, by foregrounding the everyday -- mundane -- experiences of security within communities living with and fleeing from war. We do so through three vignettes from our field research in Colombia, Lebanon and Sweden, respectively, and by highlighting the significance of ethnography for security research with communities living in regions afflicted by war. We conclude by setting out a call to action for security researchers and practitioners to consider such lived experiences in the design of security technology that aims to cater to the needs of communities in `global conflict and disaster regions'.

The Payment Protocol standard BIP70, specifying how payments in Bitcoin are performed by merchants and customers, is supported by the largest payment processors and most widely-used wallets. The protocol has been shown to be vulnerable to refund attacks due to lack of authentication of the refund addresses. In this paper, we give the first formal model of the protocol and formalise the refund address security goals for the protocol, namely refund address authentication and secrecy. The formal model utilises communication channels as abstractions conveying security goals on which the protocol modeller and verifier can rely. We analyse the Payment Protocol confirming that it is vulnerable to an attack violating the refund address authentication security goal. Moreover, we present a concrete protocol revision proposal supporting the merchant with publicly verifiable evidence that can mitigate the attack. We verify that the revised protocol meets the security goals defined for the refund address. Hence, we demonstrate that the revised protocol is secure, not only against the existing attacks, but also against any further attacks violating the formalised security goals.

Automated Insulin Delivery (AID) systems represent a significant advancement in diabetes care and wearable physiological closed-loop control technologies, integrating continuous glucose monitoring, control algorithms, and insulin pumps to improve blood glucose level control and reduce the burden of patient self-management. However, their increasing dependence on wireless communication and automatic control introduces security risks that may compromise patient privacy or result in life-threatening treatment errors. This paper presents a comprehensive survey of the AID system security landscape, covering technical vulnerabilities, regulatory frameworks, and commercial security measures. In addition, we conduct a systematic review of attack vectors and defence mechanisms proposed in the literature, following the PRISMA framework. Our findings highlight critical gaps, including the lack of specific security evaluation frameworks, insufficient protections in real-world deployments, and the need for comprehensive, lightweight, and adaptive defence mechanisms. We further investigate available research resources and outline open research challenges and future directions to guide the development of more secure and reliable AID systems. By focusing on AID systems, this review offers a representative case study for examining and improving the cybersecurity of safety-critical medical wearable systems.