arXiv:2605. 19262v2 Announce Type: replace Abstract: Masked diffusion language models (MDLMs) are emerging as a compelling new paradigm for text generation, but their training-time security remains largely unexplored.
Backdoor attacks in Large Language Models (LLMs) are a growing security concern, where models can generate adversary-chosen content. Existing defenses target backdoors one at a time and typically require knowledge of the trigger, leaving the defender at a structural disadvantage when unknown backdoors may exist in a model.
arXiv:2606. 02995v1 Announce Type: cross Abstract: Large language models remain vulnerable to jailbreak backdoor attacks, where adversaries poison safety alignment data to embed hidden triggers that bypass safety mechanisms.
The paper introduces Normal Diffusion Dynamics Learning (NDDL), a defense framework for text-to-image diffusion models that learns normal transition dynamics from benign samples. By modeling structured, timestep‑dependent patterns across cross‑attention, latent, and noise spaces, NDDL detects backdoor attacks through deviations between observed and predicted diffusion trajectories. It also localizes triggers without prior knowledge by substituting low‑semantic words, and experiments show its effectiveness across diverse attacks.
arXiv:2606. 26285v1 Announce Type: cross Abstract: Noise-based backdoor attacks on diffusion models typically rely on input-time trigger injection, untargeted activation, and out-of-distribution target generation.
Backdoor attacks pose a serious threat to the secure deployment of text-to-image (T2I) diffusion models. Existing defenses typically detect backdoors from specific abnormal patterns in internal repres...
As large language models (LLMs) are deployed in high-stakes domains, adversaries may poison training data to implant backdoors: hidden triggers that covertly manipulate model behavior at inference time. We ask whether a defender can recover such a trigger under realistic affordances, namely white-box access to the weights and knowledge of the behavior of concern, but no training data, no trusted reference model, no knowledge of the trigger, and no certainty that the model is poisoned.
Backdoor Sentinel introduces Temporal Noise Consistency (TNC), a new phenomenon where backdoor activation disrupts noise prediction consistency across adjacent diffusion timesteps, while clean inputs remain stable. Leveraging TNC, the authors propose TNC-Defense, a closed‑loop gray‑box framework that includes TNC‑Detect for auditors to identify and localize anomalous timesteps without accessing model weights, and TNC‑Detox for service providers to perform trigger‑agnostic, timestep‑aware corrections that suppress backdoor behavior. Experiments on five backdoor attacks show an 11% improvement in detection accuracy and a 98.5% invalidation rate of triggered samples with minimal impact on generation quality.
arXiv:2607. 26849v1 Announce Type: cross Abstract: As large language models (LLMs) are deployed in high-stakes domains, adversaries may poison training data to implant backdoors: hidden triggers that covertly manipulate model behavior at inference time.
arXiv:2606. 04027v1 Announce Type: cross Abstract: Diffusion large language models (dLLMs) generate text by iteratively denoising partially masked sequences under bidirectional context, exposing a safety surface distinct from autoregressive LLMs.
The paper introduces Quarantined Expert Shutdown (QES), a new backdoor containment strategy for large language models. QES allows backdoor learning to occur during training but routes it into a designated, quarantined expert that can be disabled at deployment. The method achieves significant reductions in attack success rates while largely preserving model utility.
UniGuardian is a training‑free detector for large language models that jointly identifies prompt injection, backdoor, and adversarial attacks—collectively called Prompt Trigger Attacks (PTA). It measures how structured prompt perturbations shift the model’s output distribution and uses a single‑forward strategy to detect attacks while generating text in a shared batched forward pass. Experiments show that UniGuardian accurately and efficiently identifies trigger‑activated prompts in LLMs.
arXiv:2607. 19894v1 Announce Type: cross Abstract: Large language models (LLMs) are vulnerable to backdoor attacks, where hidden triggers induce malicious outputs.