arXiv:1906.07927v4 Announce Type: cross Abstract: Deep neural networks (DNNs) have achieved great success in various applications due to their strong expressive power. However, recent studies have sh...
arXiv:2606. 11615v1 Announce Type: cross Abstract: The widespread adoption of face recognition (FR) technologies raises serious privacy concerns, as facial data can be exploited without consent.
arXiv:2607. 28936v1 Announce Type: cross Abstract: Facial biometric identification relies on the distinctiveness of user attributes within a high-dimensional embedding space.
arXiv:2605. 25194v2 Announce Type: replace Abstract: Adversarial images pose a severe security threat to multimodal large language models through prompt injection.
arXiv:2606. 24388v1 Announce Type: new Abstract: We introduce a large-scale, open-source dataset of pre-generated adversarial attacks for vision-language models (VLMs).
FSPGD introduces a feature-space black-box attack for semantic segmentation that targets intermediate representations rather than just output logits. The method uses a dual loss: an external loss to disrupt cross-model feature alignment and an internal loss to reduce consistency among same-class instances. Experiments on Pascal VOC 2012 and Cityscapes show that FSPGD outperforms existing logit-level and segmentation-specific attacks across CNN and Transformer backbones, and its adversarial examples improve robustness when used for training.
The paper investigates how Vision Language Models (VLMs) can be fooled by small, human‑imperceptible changes to images. It introduces a gradient‑based attack that targets only the vision encoder, reducing computational cost while still effectively disrupting both untargeted and targeted multimodal alignment. Experiments on open‑source VLMs such as Qwen2.5‑VL, Granite‑Vision, FastVLM, and Phi‑3.5‑Vision demonstrate that these perturbations can significantly alter the models’ textual outputs.
The paper presents three practical defense tricks to protect deep learning-based face recognition pipelines from backdoor attacks that arise when training data is sourced from potentially compromised third parties. The first trick operates during training to detect poisoned identities before deployment, while the other two work at inference time to filter suspicious inputs. Each method is described with its motivation, algorithm, computational impact, and evaluated on realistic face recognition workloads.
The paper introduces concept-based property attacks, which target logical inconsistencies in human‑interpretable concepts (e.g., “red ∧ octagonal → stop_sign”) rather than merely flipping final class labels. It formalizes these properties, proves that attacking them reduces the search space compared to class‑level attacks, and implements them using concept‑based explainable AI techniques. Empirical results across vision and code‑processing models show that retraining with these attacks preserves robustness gains while better eliminating semantic biases.
arXiv:2610.10436v1 Announce Type: new Abstract: Adversarial perturbations can alter the predictions of frozen vision-language models (VLMs) while leaving their confidence and image--text similarity p...
The paper investigates how Vision Language Models (VLMs) can be fooled by tiny, human‑imperceptible changes to images. It introduces a gradient‑based attack that targets only the vision encoder, reducing computational cost while still effectively disrupting both untargeted and targeted multimodal interpretations. Experiments on open‑source VLMs such as Qwen2.5‑VL, Granite‑Vision, FastVLM, and Phi‑3.5‑Vision demonstrate that these small perturbations can dramatically alter the models’ textual outputs.
arXiv:2601. 14300v4 Announce Type: replace Abstract: Hard-label black-box attacks, relying solely on top-1 predictions, represent one of the most challenging yet practically threat models.