arXiv subjects

Eliron Rahimi

Publications and source records attributed to Eliron Rahimi.

4 recordsLinked to original sources

arXiv2026

Breaking the Bubble: Asynchronous Pipeline Parallel Training with Bounded Weight Inconsistency

Pipeline parallelism is essential for training large neural networks, but existing schedules trade off throughput, memory, and optimization consistency. Synchronous pipelines preserve forward/backward weight consistency but suffer from bubbles; asynchronous pipelines remove bubbles but introduce weight-version mismatch, typically requiring weight stashing, prediction, or correction mechanisms. We introduce PACI (Pipeline Asynchronous training with Controlled Inconsistency), a bubble-free asynchronous pipeline method that bounds forward/backward version drift without weight stashing, prediction, additional parameter copies, or global synchronization. The key idea is to use local gradient accumulation as a version-control mechanism: by slowing parameter-version evolution relative to pipeline delay, PACI limits the number of optimizer updates crossed by any micro-batch while preserving steady-state utilization. In GPT-style language-model pretraining, PACI matches the stability and final perplexity of synchronous 1F1B-flush, retains the same peak memory footprint, achieves fully utilized pipeline throughput, and improves training time-to-accuracy by up to $1.69\times$ over the fastest flush baseline. These results show that forward/backward inconsistency need not be eliminated: when explicitly bounded, it can be safely traded for substantial efficiency gains.

cs.LG↗

arXiv2026

Step-Wise Refusal Dynamics in Autoregressive and Diffusion Language Models

Diffusion language models (DLMs) have recently emerged as a competitive alternative to autoregressive (AR) models, offering parallel decoding, competitive generation quality, and initial evidence of improved jailbreak robustness. Despite this progress, the role of sampling mechanisms in shaping refusal behavior remains poorly understood. To address this gap, we present a comprehensive study of step-wise refusal dynamics. We show that diffusion remasking can promote recovery from harmful intermediate generations, provide evidence that this behavior is tied to the sampling mechanism, and demonstrate that switching from AR to diffusion sampling improves jailbreak robustness, including under fixed model weights. To capture generation dynamics not observable at the text level, we propose the Step-Wise Refusal Internal Dynamics (SRI) signal. Consistent with our text-level findings, SRI shows that recovery fails primarily under AR sampling, with these failures often appearing anomalous relative to harmless generations in the SRI space. Based on this observation, we show that SRI enables a simple jailbreak detector that does not modify inference and generalizes to unseen attacks by training only on benign SRI signals. Our evaluation shows that this detector matches or outperforms existing jailbreak detection baselines while adding negligible overhead.

cs.LG↗

arXiv2026

Reconstructing Protected Biometric Templates from Binary Authentication Results

Biometric data is considered to be very private and highly sensitive. As such, many methods for biometric template protection were considered over the years -- from biohashing and specially crafted feature extraction procedures, to the use of cryptographic solutions such as Fuzzy Commitments or the use of Fully Homomorphic Encryption (FHE). A key question that arises is how much protection these solutions can offer when the adversary can inject samples, and observe the outputs of the system. While for systems that return the similarity score, one can use attacks such as hill-climbing, for systems where the adversary can only learn whether the authentication attempt was successful, this question remained open. In this paper, we show that it is indeed possible to reconstruct the biometric template by just observing the success/failure of the authentication attempt (given the ability to inject a sufficient amount of templates). Our attack achieves negligible template reconstruction loss and enables full recovery of facial images through a generative inversion method, forming a pipeline from binary scores to high-resolution facial images that successfully pass the system more than 98\% of the time. Our results, of course, are applicable for any protection mechanism that maintains the accuracy of the recognition.

cs.CR↗

arXiv2023

SCART: Simulation of Cyber Attacks for Real-Time

Real-Time systems are essential for promptly responding to external stimuli and completing tasks within predefined time constraints. Ensuring high reliability and robust security in these systems is therefore critical. This requires addressing reliability-related events, such as sensor failures and subsystem malfunctions, as well as cybersecurity threats. This paper introduces a novel cyber-attack simulation infrastructure designed to enhance simulation environments for real-time systems. The proposed infrastructure integrates reliability-oriented events and sophisticated cybersecurity attacks, including those targeting single or multiple sensors. We present the SCART framework and dataset, addressing a central challenge in real-time systems: the lack of scalable testing environments to assess the impact of cyber-attacks on critical systems and evaluate the effectiveness of defensive mechanisms. This limitation arises from the inherent risks of executing attacks or inducing malfunctions in operational systems. By leveraging simulation-based capabilities, the framework generates training and testing data for data-driven approaches, such as machine learning, which are otherwise difficult to train or validate under live conditions. This development enables the exploration of innovative methodologies to strengthen the resilience of real-time systems against cyber-attacks. The comprehensive functionalities of the proposed infrastructure improve the accuracy and security of critical systems while fostering the creation of advanced algorithms. These advancements hold the potential to significantly enhance anomaly detection in real-time systems and fortify their defenses against cyber threats. Our code is available at https://github.com/kfirgirstein/SCART.

cs.CR↗