BDFA: A Blind Data Adversarial Bit-flip Attack on Deep Neural Networks

Silently Disabling ECUs and Enabling Blind Attacks on the CAN Bus

The CAN Bus is crucial to the efficiency, and safety of modern vehicle infrastructure. Electronic Control Units (ECUs) exchange data across a shared bus, dropping messages whenever errors occur. If an ECU generates enough errors, their transmitter is put in a bus-off state, turning it off. Previous work abuses this process to disable ECUs, but is trivial to detect through the multiple errors transmitted over the bus. We propose a novel attack, undetectable by prior intrusion detection systems, which disables ECUs within a single message without generating any errors on the bus. Performing this…

Backdoor Attacks and Countermeasures on Deep Learning: A Comprehensive Review

This work provides the community with a timely comprehensive review of backdoor attacks and countermeasures on deep learning. According to the attacker's capability and affected stage of the machine learning pipeline, the attack surfaces are recognized to be wide and then formalized into six categorizations: code poisoning, outsourcing, pretrained, data collection, collaborative learning and post-deployment. Accordingly, attacks under each categorization are combed. The countermeasures are categorized into four general classes: blind backdoor removal, offline backdoor inspection, online backdo…

Active Fuzzing for Testing and Securing Cyber-Physical Systems

Cyber-physical systems (CPSs) in critical infrastructure face a pervasive threat from attackers, motivating research into a variety of countermeasures for securing them. Assessing the effectiveness of these countermeasures is challenging, however, as realistic benchmarks of attacks are difficult to manually construct, blindly testing is ineffective due to the enormous search spaces and resource requirements, and intelligent fuzzing approaches require impractical amounts of data and network access. In this work, we propose active fuzzing, an automatic approach for finding test suites of packet-…

(A)iSpy：用于机器学习基础设施的寄生木马

(A)iSpy: Parasitic Trojans for Machine Learning Infrastructure

研究针对机器学习基础设施的安全威胁，提出（A）iSpy寄生木马，利用执行环境信任盲点，通过主动观察和执行范式颠覆系统，能侵犯保密性、破坏完整性，可躲避扫描器，验证了其在机器学习生命周期中的扩展性及威胁实用性。