CSULoRA is a post-hoc method that corrects trained LoRA adapters by estimating a safety-aligned subspace and solving a penalized minimum-change problem to attenuate unsafe update directions while pres…

The paper introduces SORA, an adaptive adversarial training method that dynamically adjusts perturbation sizes to prevent Catastrophic Overfitting, achieving state-of-the-art robustness and clean accu…

The paper proposes WARDEN, a distributionally robust adversarial training framework that significantly reduces LLM vulnerability to adversarial attacks by dynamically reweighting hard adversarial exam…

This paper demonstrates that LLM cascade systems, designed for efficiency, are vulnerable to targeted adversarial attacks that simultaneously degrade both performance and cost-efficiency.

This survey provides a comprehensive taxonomy and vulnerability-centric analysis of adversarial attacks targeting Multimodal Large Language Models (MLLMs), offering an explanatory framework for enhanc…

DarkLLM introduces a novel framework that uses a Large Language Model (LLM) to translate natural language instructions into flexible, latent adversarial attack vectors, demonstrating a systemic vulner…

The paper proposes FedPower, a novel differentially private cross-silo Federated Learning framework that uses PowerDP to reconstruct and project client updates into a secure low-rank space, effectivel…

LoREnc is a novel, training-free framework that secures Foundation Models (FMs) and LoRA adapters against intellectual property leakage and model recovery attacks by spectrally truncating weights and…

This paper derives a method to predict and mitigate intruder dimensions caused by LoRA fine-tuning in deep learning models, improving performance and reducing forgetting.

The paper introduces TSVD, a novel framework that efficiently pre-trains LLMs by enforcing both low rank and strict weight orthonormality, achieving performance comparable to full-parameter models wit…

The paper introduces a novel Clean-Referenced Feature-Vocoder Attack, a black-box adversarial attack that perturbs high-level SSL feature representations instead of raw audio waveforms, achieving supe…

This paper provides a systematic, lifecycle-based framework for analyzing security threats and defenses across the entire fine-tuning process of LLMs, revealing that attack effectiveness is highly mod…

This paper compares the geometry of solutions left by different objectives in low-rank adapters for speech deepfake detection using the empirical Fisher diagnostic.

The paper empirically evaluates domain-adapted and general-purpose LLMs for structured threat modelling (STRIDE on 5G security), finding that domain adaptation and model size do not guarantee reliable…