Searcharxiv⌕ Search

arXiv subjects

Thomas Eisenbarth

Publications and source records attributed to Thomas Eisenbarth.

At least 19 recordsLinked to original sources
arXiv2026

Provable one-poison backdoor attacks on linear models and ReLU neural networks

Backdoor poisoning attacks are a threat to machine learning models that are trained on data collected from untrusted sources; these attacks enable attackers to inject malicious behavior into the model that can be triggered by specially crafted inputs. Prior work has established bounds on the success of backdoor attacks and their impact on the benign learning task, however, an open question is what amount of poison data is needed for a successful backdoor attack. Typical attacks either use few samples but need much information about the data points, or need to poison many data points. In this paper, we show that an adversary can mount a one-poison backdoor attack without knowledge of individual training data, requiring only coarse geometric bounds of the input space and training parameters. We identify provably sufficient conditions that allow an adversary with one poison sample with high probability to inject a backdoor into linear models and MLPs. We show that our backdoor has zero backdooring error and the injection does not significantly impact the benign learning task performance.

cs.LG↗
arXiv2026

A Comprehensive Evaluation of Code Language Models for Security Patch Detection

Automated detection of vulnerability-fixing commits (\vfcs) is critical for timely security patch deployment, as advisory databases lag patch releases by a median of 25 days and many fixes never receive advisories. Code language models are increasingly adopted for identifying VFCs, yet whether they can recognize a security fix from the code change itself remains unclear, as reported performance is shaped by commit messages, project-level data leakage, and uncertain data quality. We present a rigorous re-evaluation that jointly controls these factors through a unified framework consolidating 20 fragmented datasets spanning more than 180000 commits. Training 270 models from 125M to 80B parameters, we isolate the code signal under group-stratified code-only evaluation and assess the impact of model capacity and additional code context, ranging from intra-procedural enrichment to inter-procedural repository context. Model capacity yields clear but insufficient gains, and the evaluated context signals provide no reliable improvement under strict false positive budgets. At a false positive rate of 0.5%, every evaluated fine-tuned code-only model misses at least 80% of vulnerability fixes. A manual expert audit further shows that label error concentrates in commits lacking CVE association and primarily distorts evaluation. We derive concrete recommendations for evaluating on aggregated VFC datasets and release our unified framework and evaluation suite.

cs.SE↗
arXiv2026

FloatDoor: Platform-Triggered Backdoors in LLMs

Large language models (LLMs) are increasingly deployed in sensitive settings such as software engineering, where their outputs directly shape downstream artifacts. Recent work has shown that an identical model can produce measurably different outputs depending on the deployment platform, a consequence of non-associative floating-point arithmetic and divergent kernel implementations. We study the security implications of this platform-dependent variability and uncover a novel attack surface on LLM deployments. We introduce FloatDoor, the first input-independent, platform-triggered backdoor attack against generative LLMs. The compromised model exhibits adversary-chosen behavior when served on a target platform and is otherwise benign. FloatDoor is realized through two lightweight LoRA adapters, one that amplifies inter-platform numerical divergence and one that binds the resulting platform signature to a malicious downstream task, while leaving aggregate model utility largely intact. FloatDoor exploits a pronounced time-of-check, time-of-use gap between model auditing and serving. We demonstrate FloatDoor on Qwen3-4B across a broad range of deployment targets, including NVIDIA GPUs, Google TPUs, AWS Graviton, and Alibaba Yitian-710. As a final case study, we show that FloatDoor reliably induces exploitable code vulnerabilities on a chosen target platform. Our results establish a new class of attacks on LLM deployments and underscore the pressing need for trusted model supply chains in sensitive, LLM-powered applications.

cs.CR↗
arXiv2026

Steganography Without Modification: Hidden Communication via LLM Seeds

We demonstrate that widely deployed Large Language Model (LLM) inference stacks harbor a steganographic channel that requires no modification to model weights, sampling code, or output distributions. The channel exploits a structural property of deterministic decoding: pseudo-random number generators (PRNGs) used in inverse-transform sampling produce a seed-dependent sequence of token-level probability intervals that can be reconstructed from the generated text alone. A sender encodes a secret message in the PRNG seed before generation; a receiver reconstructs the intervals and recovers the seed, and thus the hidden payload, by exhaustive search over the seed space. We formalize two operational modes. In the known-prompt setting, sender and receiver share the prompt, enabling exact interval reconstruction and perfect seed recovery via forced alignment. In the unknown-prompt setting, only the generated text is available; approximate interval reconstruction combined with a maximum-hit-count scoring strategy still permits reliable recovery from sufficiently long outputs. Extensive experiments across six model families and five heterogeneous text domains show that, in the known-prompt setting, full 32-bit seed recovery from the complete 2^32 candidate space achieves up to 100% accuracy, depending on model and text domain, within 300 tokens and under 35 seconds on a single GPU. In the unknown-prompt setting, recovery reaches near-perfect accuracy at 600-800 tokens in about 12 seconds. We further analyze the influence of prompting strategies, tokenization ambiguities, and sampling hyperparameters on channel reliability. Moreover, we discuss several applications of our results: First, it allows for the steganographic transmission of 32 bits, but also shows that ignorance of the prompt is not a valid security assumption.

cs.CR↗
arXiv2026

PsychoSafe: Eliciting Psychologically-Informed Refusals in Large Language Models

Large language models (LLMs) routinely face requests that should be refused, creating a trade-off between helpfulness and harm prevention. However, refusals themselves can be helpful. In high-risk interactions involving crisis, coercion, or escalating intent, blunt non-compliance may prevent direct harm while still failing to support the needs of the person behind the request. We present PsychoSafe, a psychologically-informed refusal framework that reframes refusal as structured supportive communication grounded in evidence-based intervention strategies. To develop PsychoSafe, we construct a corpus of 8019 prompt-response pairs spanning five psychologically salient risk domains and apply prompting and parameter-efficient fine-tuning to Qwen 3.5 27B. On a balanced validation set of 500 prompts, evaluated with an LLM judge and validated through human ratings, PsychoSafe prompting improves overall refusal quality by 28.1% over a generic baseline, with particularly strong gains in external resource referral (+46.8%) and psychological grounding (+34.8%), while preserving downstream performance on non-refusal tasks. Fine-tuning achieves near-perfect refusal and resource-referral rates but reduces response relevance. Additional evaluations on SORRY-Bench and XSTest show strong in-domain robustness but limited out-of-domain generalization, suggesting that future work should diversify fine-tuning data to help models apply interventions selectively rather than schematically.

cs.CL↗
arXiv2026

uGen: An Agentic Framework for Generating Microarchitectural Attack PoCs

Microarchitectural attacks continue to evolve, uncovering new exploitation vectors in modern processors. From a defensive perspective, assessing a system's susceptibility to such attacks remains challenging. Developing functional attack implementations is labor-intensive, requires deep microarchitectural expertise, and is highly sensitive to execution environments. Consequently, existing attacks often lack portability, limiting systematic and scalable vulnerability assessment. Recent advances in large language models (LLMs) suggest a potential avenue for lowering these barriers. However, it remains unclear whether LLMs can reliably generate functionally correct microarchitectural attack code suitable for rigorous vulnerability testing. In this work, we present uGen, the first LLM-driven framework for automated microarchitectural attack code generation. A key challenge we address is identifying attack-specific knowledge gaps in LLMs. Through a systematic study of state-of-the-art models (GPT, Claude, and Qwen3), we find that LLMs frequently misgenerate or misplace critical attack primitives. Guided by this analysis, uGen employs a retrieval-augmented, multi-agent design that injects missing domain knowledge to synthesize functionally correct microarchitectural attack PoCs tailored to defender requirements. We evaluate uGen on cache-based and speculative-execution attacks across diverse set of microarchitectures, vulnerable functions, and LLM platforms. In the deployment stage, uGen achieves up to 100% success rate for Spectre-v1 (Claude Sonnet-4) and 80% for Prime+Probe (Qwen3-Coder). Finally, we demonstrate that uGen can generate a successful PoC code with a cost of $1.25 in under four minutes.

cs.CR↗
arXiv2026

DRAMatic Speedup: Accelerating HE Operations on a Processing-in-Memory System

Homomorphic encryption (HE) is a promising technology for confidential cloud computing, as it allows computations on encrypted data. However, HE is computationally expensive and often memory-bound on conventional computer architectures. Processing-in-Memory (PIM) is an alternative hardware architecture that integrates processing units and memory on the same chip or memory module. PIM enables higher memory bandwidth than conventional architectures and could thus be suitable for accelerating HE. We present DRAMatic, which implements operations foundational to HE on UPMEM PIM -- a programmable general-purpose PIM system developed by UPMEM. DRAMatic incorporates many arithmetic optimizations, including residue number system and number-theoretic transform techniques, and can support the large parameters required for secure homomorphic evaluations. It achieves a 334 times speed-up compared to previous HE implementations on UPMEM PIM. We also evaluate DRAMatic against Microsoft SEAL, a popular open-source HE library, regarding both runtime and energy efficiency. The results show that DRAMatic significantly closes the gap between Microsoft SEAL and HE implementations on UPMEM PIM. However, we also show that DRAMatic is currently constrained by data transfer overhead and limited multiplication performance on UPMEM PIM hardware. Finally, we discuss potential hardware extensions to UPMEM PIM.

cs.CR↗
arXiv2026

TrEEStealer: Stealing Decision Trees via Enclave Side Channels

Today, machine learning is widely applied in sensitive, security-related, and financially lucrative applications. Model extraction attacks undermine current business models where a model owner sells model access, e.g., via MLaaS APIs. Additionally, stolen models can enable powerful white-box attacks, facilitating privacy attacks on sensitive training data, and model evasion. In this paper, we focus on Decision Trees (DT), which are widely deployed in practice. Existing black-box extraction attacks for DTs are either query-intensive, make strong assumptions about the DT structure, or rely on rich API information. To limit attacks to the black-box setting, CPU vendors introduced Trusted Execution Environments (TEE) that use hardware-mechanisms to isolate workloads from external parties, e.g., MLaaS providers. We introduce TrEEStealer, a high-fidelity extraction attack for stealing TEE-protected DTs. TrEEStealer exploits TEE-specific side-channels to steal DTs efficiently and without strong assumptions about the API output or DT structure. The extraction efficacy stems from a novel algorithm that maximizes the information derived from each query by coupling Control-Flow Information (CFI) with passive information tracking. We use two primitives to acquire CFI: for AMD SEV, we follow previous work using the SEV-Step framework and performance counters. For Intel SGX, we reproduce prior findings on current Xeon 6 CPUs and construct a new primitive to efficiently extract the branch history of inference runs through the Branch-History-Register. We found corresponding vulnerabilities in three popular libraries: OpenCV, mlpack, and emlearn. We show that TrEEStealer achieves superior efficiency and extraction fidelity compared to prior attacks. Our work establishes a new state-of-the-art for DT extraction and confirms that TEEs fail to protect against control-flow leakage.

cs.CR↗
arXiv2026

HPCCFA: Leveraging Hardware Performance Counters for Control Flow Attestation

Trusted Execution Environments (TEEs) allow the secure execution of code on remote systems without the need to trust their operators. They use static attestation as a central mechanism for establishing trust, allowing remote parties to verify that their code is executed unmodified in an isolated environment. However, this form of attestation does not cover runtime attacks, where an attacker exploits vulnerabilities in the software inside the TEE. Control Flow Attestation (CFA), a form of runtime attestation, is designed to detect such attacks. In this work, we present a method to extend TEEs with CFA and discuss how it can prevent exploitation in the event of detected control flow violations. Furthermore, we introduce HPCCFA, a mechanism that uses HPCs for CFA purposes, enabling hardware-backed trace generation on commodity CPUs. We demonstrate the feasibility of HPCCFA on a proof-of-concept implementation for Keystone on RISC-V. Our evaluation investigates the interplay of the number of measurement points and runtime protection, and reveals a trade-off between detection reliability and performance overhead.

cs.CR↗
arXiv2026

Coverage-Guided Multi-Agent Harness Generation for Java Library Fuzzing

Coverage-guided fuzzing has proven effective for software testing, but targeting library code requires specialized fuzz harnesses that translate fuzzer-generated inputs into valid API invocations. Manual harness creation is time-consuming and requires deep understanding of API semantics, initialization sequences, and exception handling contracts. We present a multi-agent architecture that automates fuzz harness generation for Java libraries through specialized LLM-powered agents. Five ReAct agents decompose the workflow into research, synthesis, compilation repair, coverage analysis, and refinement. Rather than preprocessing entire codebases, agents query documentation, source code, and callgraph information on demand through the Model Context Protocol, maintaining focused context while exploring complex dependencies. To enable effective refinement, we introduce method-targeted coverage that tracks coverage only during target method execution to isolate target behavior, and agent-guided termination that examines uncovered source code to distinguish productive refinement opportunities from diminishing returns. We evaluated our approach on seven target methods from six widely-deployed Java libraries totaling 115,000+ Maven dependents. Our generated harnesses achieve a median 26\% improvement over OSS-Fuzz baselines and outperform Jazzer AutoFuzz by 5\% in package-scope coverage. Generation costs average \$3.20 and 10 minutes per harness, making the approach practical for continuous fuzzing workflows. During a 12-hour fuzzing campaign, our generated harnesses discovered 3 bugs in projects that are already integrated into OSS-Fuzz, demonstrating the effectiveness of the generated harnesses.

cs.SE↗
arXiv2026

SLasH-DSA: Breaking SLH-DSA Using an Extensible End-To-End Rowhammer Framework

As quantum computing advances, Post-Quantum Cryptography (PQC) schemes are adopted to replace classical algorithms. Among them is the Stateless Hash-Based Digital Signature Algorithm (SLH-DSA) that was recently standardized by NIST and is favored for its conservative security basis. In this work, we present the first software-only universal forgery attack on SLH-DSA, leveraging Rowhammer-induced bit flips to corrupt the internal state and forge signatures. While prior work targeted embedded systems and required physical access, our attack is software-only, targeting commodity desktop and server hardware, significantly broadening the threat model. We demonstrate full end-to-end attacks against SLH-DSA in OpenSSL 3.5.1, achieving universal forgery for the SHAKE-128f (deterministic), SHA2-128s, and SHAKE-192f (randomized) parameter sets after one hour (deterministic) or eight hours (randomized) of hammering and post-processing ranging from minutes to an hour, and showing theoretical attack complexities for most parameter sets. Our post-processing is informed by a novel complexity analysis that, given a concrete set of faulty signatures, identifies the most promising computational path to pursue. To enable the attack, we introduce Swage, a modular and extensible framework for implementing end-to-end Rowhammer-based fault attacks. Swage abstracts and automates key components of practical Rowhammer attacks. Unlike prior tooling, Swage is untangled from the attacked code, making it reusable and suitable for frictionless analysis of different targets. Our findings highlight that even theoretically sound PQC schemes can fail under real-world conditions, underscoring the need for additional implementation hardening or hardware defenses against Rowhammer.

cs.CR↗
arXiv2026

Beyond Accuracy: Characterizing Code Comprehension Capabilities in (Large) Language Models

Large Language Models (LLMs) are increasingly integrated into software engineering workflows, yet current benchmarks provide only coarse performance summaries that obscure the diverse capabilities and limitations of these models. This paper investigates whether LLMs' code-comprehension performance aligns with traditional human-centric software metrics or instead reflects distinct, non-human regularities. We introduce a diagnostic framework that reframes code understanding as a binary input-output consistency task, enabling the evaluation of classification and generative models. Using a large-scale dataset, we correlate model performance with traditional, human-centric complexity metrics, such as lexical size, control-flow complexity, and abstract syntax tree structure. Our analyses reveal minimal correlation between human-defined metrics and LLM success (AUROC 0.63), while shadow models achieve substantially higher predictive performance (AUROC 0.86), capturing complex, partially predictable patterns beyond traditional software measures. These findings suggest that LLM comprehension reflects model-specific regularities only partially accessible through either human-designed or learned features, emphasizing the need for benchmark methodologies that move beyond aggregate accuracy and toward instance-level diagnostics, while acknowledging fundamental limits in predicting correct outcomes.

cs.SE↗
arXiv2025

Trace Gadgets: Minimizing Code Context for Machine Learning-Based Vulnerability Prediction

As the number of web applications and API endpoints exposed to the Internet continues to grow, so does the number of exploitable vulnerabilities. Manually identifying such vulnerabilities is tedious. Meanwhile, static security scanners tend to produce many false positives. While machine learning-based approaches are promising, they typically perform well only in scenarios where training and test data are closely related. A key challenge for ML-based vulnerability detection is providing suitable and concise code context, as excessively long contexts negatively affect the code comprehension capabilities of machine learning models, particularly smaller ones. This work introduces Trace Gadgets, a novel code representation that minimizes code context by removing non-related code. Trace Gadgets precisely capture the statements that cover the path to the vulnerability. As input for ML models, Trace Gadgets provide a minimal but complete context, thereby improving the detection performance. Moreover, we collect a large-scale dataset generated from real-world applications with manually curated labels to further improve the performance of ML-based vulnerability detectors. Our results show that state-of-the-art machine learning models perform best when using Trace Gadgets compared to previous code representations, surpassing the detection capabilities of industry-standard static scanners such as GitHub's CodeQL by at least 4% on a fully unseen dataset. By applying our framework to real-world applications, we identify and report previously unknown vulnerabilities in widely deployed software.

cs.CR↗
arXiv2025

BarkBeetle: Stealing Decision Tree Models with Fault Injection

Machine learning models, particularly decision trees (DTs), are widely adopted across various domains due to their interpretability and efficiency. However, as ML models become increasingly integrated into privacy-sensitive applications, concerns about their confidentiality have grown, particularly in light of emerging threats such as model extraction and fault injection attacks. Assessing the vulnerability of DTs under such attacks is therefore important. In this work, we present BarkBeetle, a novel attack that leverages fault injection to extract internal structural information of DT models. BarkBeetle employs a bottom-up recovery strategy that uses targeted fault injection at specific nodes to efficiently infer feature splits and threshold values. Our proof-of-concept implementation demonstrates that BarkBeetle requires significantly fewer queries and recovers more structural information compared to prior approaches, when evaluated on DTs trained with public UCI datasets. To validate its practical feasibility, we implement BarkBeetle on a Raspberry Pi RP2350 board and perform fault injections using the Faultier voltage glitching tool. As BarkBeetle targets general DT models, we also provide an in-depth discussion on its applicability to a broader range of tree-based applications, including data stream classification, DT variants, and cryptography schemes.

cs.CR↗
arXiv2025

Obelix: Mitigating Side-Channels Through Dynamic Obfuscation

Trusted execution environments (TEEs) offer hardware-assisted means to protect code and data. However, as shown in numerous results over the years, attackers can use side-channels to leak data access patterns and even single-step the code. While the vendors are slowly introducing hardware-based countermeasures for some attacks, others will stay unaddressed. This makes a software-level countermeasure desirable, but current available solutions only address very specific attack vectors or have a narrow leakage model. In this work, we take a holistic view at the vulnerabilities of TEEs and design a tool named Obelix, which is the first to protect both code and data against a wide range of TEE attacks, from cache attacks over single-stepping to ciphertext side-channels. We analyze the practically achievable precision of state-of-the-art single-stepping tools, and present an algorithm which uses that knowledge to divide a program into uniform code blocks, that are indistinguishable for a strong attacker. By storing these blocks and the program data in oblivious RAM, the attacker cannot follow execution, effectively protecting both secret code and data. We describe how we automate our approach to make it available for developers who are unfamiliar with side-channels. As an obfuscation tool, Obelix comes with a considerable performance overhead, but compensates this with strong security guarantees and easy applicability without requiring any expert knowledge.

cs.CR↗
arXiv2025

Silenzio: Secure Non-Interactive Outsourced MLP Training

Outsourcing ML training to cloud-service-providers presents a compelling opportunity for resource constrained clients, while it simultaneously bears inherent privacy risks. We introduce Silenuio, the first fully non-interactive outsourcing scheme for the training of MLPs that achieves 128bit security using FHE (precisely TFHE). Unlike traditional MPC-based protocols that necessitate interactive communication between the client and server(s) or non-collusion assumptions among multiple servers, Silenzio enables the "fire-and-forget" paradigm without such assumptions. In this approach, the client encrypts the training data once, and the server performs the training without any further interaction. Silenzio operates entirely over low-bitwidth integer to mitigate the computational overhead inherent to FHE. Our approach features a novel low-bitwidth matrix multiplication gadget that leverages input-dependent residue number systems, ensuring that no intermediate value overflows 8bit. Starting from an RNS-to-MRNS conversion process, we propose an efficient block-scaling mechanism, which approximately shifts encrypted tensor values to their user-specified most significant bits. To instantiate the backpropagation of the error, Silenzio introduces a low-bitwidth gradient computation for the cross-entropy loss. We evaluate Silenzio on standard MLP training tasks regarding runtime as well as model performance and achieve similar classification accuracy as MLPs trained using PyTorch with 32bit floating-point computations. Our open-source implementation of Silenzio represents a significant advancement in privacy-preserving ML, providing a new baseline for secure and non-interactive outsourced MLP training.

cs.CR↗
arXiv2025

Prompt Pirates Need a Map: Stealing Seeds helps Stealing Prompts

Diffusion models have significantly advanced text-to-image generation, enabling the creation of highly realistic images conditioned on textual prompts and seeds. Given the considerable intellectual and economic value embedded in such prompts, prompt theft poses a critical security and privacy concern. In this paper, we investigate prompt-stealing attacks targeting diffusion models. We reveal that numerical optimization-based prompt recovery methods are fundamentally limited as they do not account for the initial random noise used during image generation. We identify and exploit a noise-generation vulnerability (CWE-339), prevalent in major image-generation frameworks, originating from PyTorch's restriction of seed values to a range of $2^{32}$ when generating the initial random noise on CPUs. Through a large-scale empirical analysis conducted on images shared via the popular platform CivitAI, we demonstrate that approximately 95% of these images' seed values can be effectively brute-forced in 140 minutes per seed using our seed-recovery tool, SeedSnitch. Leveraging the recovered seed, we propose PromptPirate, a genetic algorithm-based optimization method explicitly designed for prompt stealing. PromptPirate surpasses state-of-the-art methods, i.e., PromptStealer, P2HP, and CLIP-Interrogator, achieving an 8-11% improvement in LPIPS similarity. Furthermore, we introduce straightforward and effective countermeasures that render seed stealing, and thus optimization-based prompt stealing, ineffective. We have disclosed our findings responsibly and initiated coordinated mitigation efforts with the developers to address this critical vulnerability.

cs.CR↗
arXiv2025

AutoStub: Genetic Programming-Based Stub Creation for Symbolic Execution

Symbolic execution is a powerful technique for software testing, but suffers from limitations when encountering external functions, such as native methods or third-party libraries. Existing solutions often require additional context, expensive SMT solvers, or manual intervention to approximate these functions through symbolic stubs. In this work, we propose a novel approach to automatically generate symbolic stubs for external functions during symbolic execution that leverages Genetic Programming. When the symbolic executor encounters an external function, AutoStub generates training data by executing the function on randomly generated inputs and collecting the outputs. Genetic Programming then derives expressions that approximate the behavior of the function, serving as symbolic stubs. These automatically generated stubs allow the symbolic executor to continue the analysis without manual intervention, enabling the exploration of program paths that were previously intractable. We demonstrate that AutoStub can automatically approximate external functions with over 90% accuracy for 55% of the functions evaluated, and can infer language-specific behaviors that reveal edge cases crucial for software testing.

cs.SE↗