arXiv subjects

Robert Bridges

Publications and source records attributed to Robert Bridges.

4 recordsLinked to original sources

arXiv2026

Prefilling the Reasoning Channel: Output-Prefix Attacks on Reasoning LLMs

Large Language Models (LLMs) consume and produce a single sequence of text; hence, if text can be added to the beginning of the LLM's response, i.e., an output prefix, then all subsequent tokens will be conditioned on it. This output-prefix attack technique is a cheap black-box prompt injection. Prior work has shown this type of attack can reliably jailbreak non-reasoning models. Most reasoning models add an intermediate scratchpad reasoning step before the assistant's final response. The ability to edit this reasoning channel is exposed by some APIs and attack vectors can be leveraged for reasoning injection attacks. We present the first systematic, controlled study that isolates the scratchpad reasoning channel as an output-prefix attack vector, and the first to compare reasoning-only, output-prefix-only and reasoning-plus-output-prefix attacks across both exposed- and hidden-reasoning models. Using a factorial design of 3 prefix types $\times$ 2 reasoning injections over $1{,}800$ test cases drawn from AdvBench, we attack three 2026-era frontier models Gemini 3 Flash Preview, DeepSeek V4 Flash, and Claude Haiku 4.5. We find that injecting malicious reasoning alone is essentially inert ($\approx0\%$ attack success), but injecting the same reasoning together with a trivial output prefix raises the attack success rate to as high as $99\%$ for some models. For this type of attack we find that contextual prefixes work better than static prefixes; and that susceptibility is dependent on the model.

cs.CR↗

arXiv2026

A Nonlinear Singular Value Theory for Neural Networks

Recently Brown et al. [2025] established a singular value decomposition (SVD) for maps (especially nonlinear) satisfying certain norm conditions. We prove that most modern neural architectures admit this nonlinear SVD (NLSVD) representation---with no change in input--output behavior---and enumerate the classes covered. In this factorization the network is a left-invertible nonlinear map followed by a final linear layer. Moreover, the left-invertible factor is norm-preserving, so distances in the embedding (activations before the final linear layer) calibrate directly to distances in input space. We introduce a flexible architecture that yields an explicit decomposition at training time, a data-driven algorithm for estimating the representation from trained models, and the mathematical foundations for nonlinear analogues of row and null spaces in neural networks. Empirical case studies illustrate uses of the theory for latent-space pullback (visualization and data generation), bias detection, and membership-inference robustness under training. Altogether, these foundations support new approaches to core problems in neural-network analysis.

cs.LG↗

arXiv2020

An Assessment of the Usability of Machine Learning Based Tools for the Security Operations Center

Gartner, a large research and advisory company, anticipates that by 2024 80% of security operation centers (SOCs) will use machine learning (ML) based solutions to enhance their operations. In light of such widespread adoption, it is vital for the research community to identify and address usability concerns. This work presents the results of the first in situ usability assessment of ML-based tools. With the support of the US Navy, we leveraged the national cyber range, a large, air-gapped cyber testbed equipped with state-of-the-art network and user emulation capabilities, to study six US Naval SOC analysts' usage of two tools. Our analysis identified several serious usability issues, including multiple violations of established usability heuristics form user interface design. We also discovered that analysts lacked a clear mental model of how these tools generate scores, resulting in mistrust and/or misuse of the tools themselves. Surprisingly, we found no correlation between analysts' level of education or years of experience and their performance with either tool, suggesting that other factors such as prior background knowledge or personality play a significant role in ML-based tool usage. Our findings demonstrate that ML-based security tool vendors must put a renewed focus on working with analysts, both experienced and inexperienced, to ensure that their systems are usable and useful in real-world security operations settings.

cs.HC↗

arXiv2018

Towards Malware Detection via CPU Power Consumption: Data Collection Design and Analytics (Extended Version)

This paper presents an experimental design and data analytics approach aimed at power-based malware detection on general-purpose computers. Leveraging the fact that malware executions must consume power, we explore the postulate that malware can be accurately detected via power data analytics. Our experimental design and implementation allow for programmatic collection of CPU power profiles for fixed tasks during uninfected and infected states using five different rootkits. To characterize the power consumption profiles, we use both simple statistical and novel, sophisticated features. We test a one-class anomaly detection ensemble (that baselines non-infected power profiles) and several kernel-based SVM classifiers (that train on both uninfected and infected profiles) in detecting previously unseen malware and clean profiles. The anomaly detection system exhibits perfect detection when using all features and tasks, with smaller false detection rate than the supervised classifiers. The primary contribution is the proof of concept that baselining power of fixed tasks can provide accurate detection of rootkits. Moreover, our treatment presents engineering hurdles needed for experimentation and allows analysis of each statistical feature individually. This work appears to be the first step towards a viable power-based detection capability for general-purpose computers, and presents next steps toward this goal.

cs.CR↗