Jailbreaking Multimodal Large Language Models via Shuffle Inconsistency

Multimodal Large Language Models (MLLMs) have achieved impressive performance and have been put into practical use in commercial applications, but they still have potential safety mechanism vulnerabilities. Jailbreak attacks are red teaming methods that aim to bypass safety mechanisms and discover MLLMs' potential risks. Existing MLLMs' jailbreak methods often bypass the model's safety mechanism through complex optimization methods or carefully designed image and text prompts. Despite achieving some progress, they have a low attack success rate on commercial closed-source MLLMs. Unlike previous research, we empirically find that there exists a Shuffle Inconsistency between MLLMs' comprehension ability and safety ability for the shuffled harmful instruction. That is, from the perspective of comprehension ability, MLLMs can understand the shuffled harmful text-image instructions well. However, they can be easily bypassed by the shuffled harmful instructions from the perspective of safety ability, leading to harmful responses. Then we innovatively propose a text-image jailbreak attack named SI-Attack. Specifically, to fully utilize the Shuffle Inconsistency and overcome the shuffle randomness, we apply a query-based black-box optimization method to select the most harmful shuffled inputs based on the feedback of the toxic judge model. A series of experiments show that SI-Attack can improve the attack's performance on three benchmarks. In particular, SI-Attack can obviously improve the attack success rate for commercial MLLMs such as GPT-4o or Claude-3.5-Sonnet.

OODFace: Benchmarking Robustness of Face Recognition under Common Corruptions and Appearance Variations

With the rise of deep learning, facial recognition technology has seen extensive research and rapid development. Although facial recognition is considered a mature technology, we find that existing open-source models and commercial algorithms lack robustness in certain real-world Out-of-Distribution (OOD) scenarios, raising concerns about the reliability of these systems. In this paper, we introduce OODFace, which explores the OOD challenges faced by facial recognition models from two perspectives: common corruptions and appearance variations. We systematically design 30 OOD scenarios across 9 major categories tailored for facial recognition. By simulating these challenges on public datasets, we establish three robustness benchmarks: LFW-C/V, CFP-FP-C/V, and YTF-C/V. We then conduct extensive experiments on 19 different facial recognition models and 3 commercial APIs, along with extended experiments on face masks, Vision-Language Models (VLMs), and defense strategies to assess their robustness. Based on the results, we draw several key insights, highlighting the vulnerability of facial recognition systems to OOD data and suggesting possible solutions. Additionally, we offer a unified toolkit that includes all corruption and variation types, easily extendable to other datasets. We hope that our benchmarks and findings can provide guidance for future improvements in facial recognition model robustness.

Improving Adversarial Robust Fairness via Anti-Bias Soft Label Distillation

Adversarial Training (AT) has been widely proved to be an effective method to improve the adversarial robustness against adversarial examples for Deep Neural Networks (DNNs). As a variant of AT, Adversarial Robustness Distillation (ARD) has demonstrated its superior performance in improving the robustness of small student models with the guidance of large teacher models. However, both AT and ARD encounter the robust fairness problem: these models exhibit strong robustness when facing part of classes (easy class), but weak robustness when facing others (hard class). In this paper, we give an in-depth analysis of the potential factors and argue that the smoothness degree of samples' soft labels for different classes (i.e., hard class or easy class) will affect the robust fairness of DNN models from both empirical observation and theoretical analysis. Based on the above finding, we propose an Anti-Bias Soft Label Distillation (ABSLD) method to mitigate the adversarial robust fairness problem within the framework of Knowledge Distillation (KD). Specifically, ABSLD adaptively reduces the student's error risk gap between different classes to achieve fairness by adjusting the class-wise smoothness degree of samples' soft labels during the training process, and the smoothness degree of soft labels is controlled by assigning different temperatures in KD to different classes. Extensive experiments demonstrate that ABSLD outperforms state-of-the-art AT, ARD, and robust fairness methods in terms of overall performance of robustness and fairness.

Mitigating the Accuracy-Robustness Trade-off via Multi-Teacher Adversarial Distillation

Adversarial training is a practical approach for improving the robustness of deep neural networks against adversarial attacks. Although bringing reliable robustness, the performance toward clean examples is negatively affected after adversarial training, which means a trade-off exists between accuracy and robustness. Recently, some studies have tried to use knowledge distillation methods in adversarial training, achieving competitive performance in improving the robustness but the accuracy for clean samples is still limited. In this paper, to mitigate the accuracy-robustness trade-off, we introduce the Multi-Teacher Adversarial Robustness Distillation (MTARD) to guide the model's adversarial training process by applying a strong clean teacher and a strong robust teacher to handle the clean examples and adversarial examples, respectively. During the optimization process, to ensure that different teachers show similar knowledge scales, we design the Entropy-Based Balance algorithm to adjust the teacher's temperature and keep the teachers' information entropy consistent. Besides, to ensure that the student has a relatively consistent learning speed from multiple teachers, we propose the Normalization Loss Balance algorithm to adjust the learning weights of different types of knowledge. A series of experiments conducted on public datasets demonstrate that MTARD outperforms the state-of-the-art adversarial training and distillation methods against various adversarial attacks.

Boosting Adversarial Transferability with Learnable Patch-wise Masks

Adversarial examples have raised widespread attention in security-critical applications because of their transferability across different models. Although many methods have been proposed to boost adversarial transferability, a gap still exists in the practical demand. In this paper, we argue that the model-specific discriminative regions are a key factor to cause the over-fitting to the source model, and thus reduce the transferability to the target model. For that, a patch-wise mask is utilized to prune the model-specific regions when calculating adversarial perturbations. To accurately localize these regions, we present a learnable approach to optimize the mask automatically. Specifically, we simulate the target models in our framework, and adjust the patch-wise mask according to the feedback of simulated models. To improve the efficiency, Differential Evolutionary (DE) algorithm is utilized to search for patch-wise masks for a specific image. During iterative attacks, the learned masks are applied to the image to drop out the patches related to model-specific regions, thus making the gradients more generic and improving the adversarial transferability. The proposed approach is a pre-processing method and can be integrated with existing gradient-based methods to further boost the transfer attack success rate. Extensive experiments on the ImageNet dataset demonstrate the effectiveness of our method. We incorporate the proposed approach with existing methods in the ensemble attacks and achieve an average success rate of 93.01% against seven advanced defense methods, which can effectively enhance the state-of-the-art transfer-based attack performance.

Revisiting the Trade-off between Accuracy and Robustness via Weight Distribution of Filters

Adversarial attacks have been proven to be potential threats to Deep Neural Networks (DNNs), and many methods are proposed to defend against adversarial attacks. However, while enhancing the robustness, the clean accuracy will decline to a certain extent, implying a trade-off existed between the accuracy and robustness. In this paper, we firstly empirically find an obvious distinction between standard and robust models in the filters' weight distribution of the same architecture, and then theoretically explain this phenomenon in terms of the gradient regularization, which shows this difference is an intrinsic property for DNNs, and thus a static network architecture is difficult to improve the accuracy and robustness at the same time. Secondly, based on this observation, we propose a sample-wise dynamic network architecture named Adversarial Weight-Varied Network (AW-Net), which focuses on dealing with clean and adversarial examples with a ``divide and rule" weight strategy. The AW-Net dynamically adjusts network's weights based on regulation signals generated by an adversarial detector, which is directly influenced by the input sample. Benefiting from the dynamic network architecture, clean and adversarial examples can be processed with different network weights, which provides the potentiality to enhance the accuracy and robustness simultaneously. A series of experiments demonstrate that our AW-Net is architecture-friendly to handle both clean and adversarial examples and can achieve better trade-off performance than state-of-the-art robust models.

MedLocker: A Transferable Adversarial Watermarking for Preventing Unauthorized Analysis of Medical Image Dataset

The collection of medical image datasets is a demanding and laborious process that requires significant resources. Furthermore, these medical datasets may contain personally identifiable information, necessitating measures to ensure that unauthorized access is prevented. Failure to do so could violate the intellectual property rights of the dataset owner and potentially compromise the privacy of patients. As a result, safeguarding medical datasets and preventing unauthorized usage by AI diagnostic models is a pressing challenge. To address this challenge, we propose a novel visible adversarial watermarking method for medical image copyright protection, called MedLocker. Our approach involves continuously optimizing the position and transparency of a watermark logo, which reduces the performance of the target model, leading to incorrect predictions. Importantly, we ensure that our method minimizes the impact on clinical visualization by constraining watermark positions using semantical masks (WSM), which are bounding boxes of lesion regions based on semantic segmentation. To ensure the transferability of the watermark across different models, we verify the cross-model transferability of the watermark generated on a single model. Additionally, we generate a unique watermark parameter list each time, which can be used as a certification to verify the authorization. We evaluate the performance of MedLocker on various mainstream backbones and validate the feasibility of adversarial watermarking for copyright protection on two widely-used diabetic retinopathy detection datasets. Our results demonstrate that MedLocker can effectively protect the copyright of medical datasets and prevent unauthorized users from analyzing medical images with AI diagnostic models.