Zhengyu Zhao

Yiyong Liu, Zhengyu Zhao, Michael Backes et al.

Machine learning models are vulnerable to membership inference attacks in which an adversary aims to predict whether or not a particular sample was contained in the target model's training dataset. Existing attack methods have commonly exploited the output information (mostly, losses) solely from the given target model. As a result, in practical scenarios where both the member and non-member samples yield similarly small losses, these methods are naturally unable to differentiate between them. To address this limitation, in this paper, we propose a new attack method, called \system, which can exploit the membership information from the whole training process of the target model for improving the attack performance. To mount the attack in the common black-box setting, we leverage knowledge distillation, and represent the membership information by the losses evaluated on a sequence of intermediate models at different distillation epochs, namely \emph{distilled loss trajectory}, together with the loss from the given target model. Experimental results over different datasets and model architectures demonstrate the great advantage of our attack in terms of different metrics. For example, on CINIC-10, our attack achieves at least 6$\times$ higher true-positive rate at a low false-positive rate of 0.1\% than existing methods. Further analysis demonstrates the general effectiveness of our attack in more strict scenarios.

Songlong Xing, Weijie Wang, Zhengyu Zhao et al.

Despite their impressive zero-shot abilities, vision-language models such as CLIP have been shown to be susceptible to adversarial attacks. To enhance its adversarial robustness, recent studies finetune the pretrained vision encoder of CLIP with adversarial examples on a proxy dataset such as ImageNet by aligning adversarial images with correct class labels. However, these methods overlook the important roles of training data distributions and learning objectives, resulting in reduced zero-shot capabilities and limited transferability of robustness across domains and datasets. In this work, we propose a simple yet effective paradigm AdvFLYP, which follows the training recipe of CLIP's pretraining process when performing adversarial finetuning to the model. Specifically, AdvFLYP finetunes CLIP with adversarial images created based on image-text pairs collected from the web, and match them with their corresponding texts via a contrastive loss. To alleviate distortion of adversarial image embeddings of noisy web images, we further propose to regularise AdvFLYP by penalising deviation of adversarial image features. We show that logit- and feature-level regularisation terms benefit robustness and clean accuracy, respectively. Extensive experiments on 14 downstream datasets spanning various domains show the superiority of our paradigm over mainstream practices. Our code and model weights are released at https://github.com/Sxing2/AdvFLYP.

Yihan Ma, Zhengyu Zhao, Xinlei He et al.

Large text-to-image models have shown remarkable performance in synthesizing high-quality images. In particular, the subject-driven model makes it possible to personalize the image synthesis for a specific subject, e.g., a human face or an artistic style, by fine-tuning the generic text-to-image model with a few images from that subject. Nevertheless, misuse of subject-driven image synthesis may violate the authority of subject owners. For example, malicious users may use subject-driven synthesis to mimic specific artistic styles or to create fake facial images without authorization. To protect subject owners against such misuse, recent attempts have commonly relied on adversarial examples to indiscriminately disrupt subject-driven image synthesis. However, this essentially prevents any benign use of subject-driven synthesis based on protected images. In this paper, we take a different angle and aim at protection without sacrificing the utility of protected images for general synthesis purposes. Specifically, we propose GenWatermark, a novel watermark system based on jointly learning a watermark generator and a detector. In particular, to help the watermark survive the subject-driven synthesis, we incorporate the synthesis process in learning GenWatermark by fine-tuning the detector with synthesized images for a specific subject. This operation is shown to largely improve the watermark detection accuracy and also ensure the uniqueness of the watermark for each individual subject. Extensive experiments validate the effectiveness of GenWatermark, especially in practical scenarios with unknown models and text prompts (74% Acc.), as well as partial data watermarking (80% Acc. for 1/4 watermarking). We also demonstrate the robustness of GenWatermark to two potential countermeasures that substantially degrade the synthesis quality.

Zhengyu Zhao, Hanwei Zhang, Renjue Li et al.

Transfer adversarial attacks raise critical security concerns in real-world, black-box scenarios. However, the actual progress of this field is difficult to assess due to two common limitations in existing evaluations. First, different methods are often not systematically and fairly evaluated in a one-to-one comparison. Second, only transferability is evaluated but another key attack property, stealthiness, is largely overlooked. In this work, we design good practices to address these limitations, and we present the first comprehensive evaluation of transfer attacks, covering 23 representative attacks against 9 defenses on ImageNet. In particular, we propose to categorize existing attacks into five categories, which enables our systematic category-wise analyses. These analyses lead to new findings that even challenge existing knowledge and also help determine the optimal attack hyperparameters for our attack-wise comprehensive evaluation. We also pay particular attention to stealthiness, by adopting diverse imperceptibility metrics and looking into new, finer-grained characteristics. Overall, our new insights into transferability and stealthiness lead to actionable good practices for future evaluations.

Weipeng Jiang, Zhenting Wang, Juan Zhai et al.

Despite prior safety alignment efforts, mainstream LLMs can still generate harmful and unethical content when subjected to jailbreaking attacks. Existing jailbreaking methods fall into two main categories: template-based and optimization-based methods. The former requires significant manual effort and domain knowledge, while the latter, exemplified by Greedy Coordinate Gradient (GCG), which seeks to maximize the likelihood of harmful LLM outputs through token-level optimization, also encounters several limitations: requiring white-box access, necessitating pre-constructed affirmative phrase, and suffering from low efficiency. In this paper, we present ECLIPSE, a novel and efficient black-box jailbreaking method utilizing optimizable suffixes. Drawing inspiration from LLMs' powerful generation and optimization capabilities, we employ task prompts to translate jailbreaking goals into natural language instructions. This guides the LLM to generate adversarial suffixes for malicious queries. In particular, a harmfulness scorer provides continuous feedback, enabling LLM self-reflection and iterative optimization to autonomously and efficiently produce effective suffixes. Experimental results demonstrate that ECLIPSE achieves an average attack success rate (ASR) of 0.92 across three open-source LLMs and GPT-3.5-Turbo, significantly surpassing GCG in 2.4 times. Moreover, ECLIPSE is on par with template-based methods in ASR while offering superior attack efficiency, reducing the average attack overhead by 83%.

Zhuoran Liu, Zhengyu Zhao, Martha Larson

Perturbative availability poisons (PAPs) add small changes to images to prevent their use for model training. Current research adopts the belief that practical and effective approaches to countering PAPs do not exist. In this paper, we argue that it is time to abandon this belief. We present extensive experiments showing that 12 state-of-the-art PAP methods are vulnerable to Image Shortcut Squeezing (ISS), which is based on simple compression. For example, on average, ISS restores the CIFAR-10 model accuracy to $81.73\%$, surpassing the previous best preprocessing-based countermeasures by $37.97\%$ absolute. ISS also (slightly) outperforms adversarial training and has higher generalizability to unseen perturbation norms and also higher efficiency. Our investigation reveals that the property of PAP perturbations depends on the type of surrogate model used for poison generation, and it explains why a specific ISS compression yields the best performance for a specific type of PAP perturbation. We further test stronger, adaptive poisoning, and show it falls short of being an ideal defense against ISS. Overall, our results demonstrate the importance of considering various (simple) countermeasures to ensure the meaningfulness of analysis carried out during the development of PAP methods.

Hai Huang, Zhengyu Zhao, Michael Backes et al.

Large language models (LLMs) have demonstrated superior performance compared to previous methods on various tasks, and often serve as the foundation models for many researches and services. However, the untrustworthy third-party LLMs may covertly introduce vulnerabilities for downstream tasks. In this paper, we explore the vulnerability of LLMs through the lens of backdoor attacks. Different from existing backdoor attacks against LLMs, ours scatters multiple trigger keys in different prompt components. Such a Composite Backdoor Attack (CBA) is shown to be stealthier than implanting the same multiple trigger keys in only a single component. CBA ensures that the backdoor is activated only when all trigger keys appear. Our experiments demonstrate that CBA is effective in both natural language processing (NLP) and multimodal tasks. For instance, with $3\%$ poisoning samples against the LLaMA-7B model on the Emotion dataset, our attack achieves a $100\%$ Attack Success Rate (ASR) with a False Triggered Rate (FTR) below $2.06\%$ and negligible model accuracy degradation. Our work highlights the necessity of increased security research on the trustworthiness of foundation LLMs.

Zhengyu Zhao, Hanwei Zhang, Renjue Li et al.

Transferable adversarial images raise critical security concerns for computer vision systems in real-world, black-box attack scenarios. Although many transfer attacks have been proposed, existing research lacks a systematic and comprehensive evaluation. In this paper, we systemize transfer attacks into five categories around the general machine learning pipeline and provide the first comprehensive evaluation, with 23 representative attacks against 11 representative defenses, including the recent, transfer-oriented defense and the real-world Google Cloud Vision. In particular, we identify two main problems of existing evaluations: (1) for attack transferability, lack of intra-category analyses with fair hyperparameter settings, and (2) for attack stealthiness, lack of diverse measures. Our evaluation results validate that these problems have indeed caused misleading conclusions and missing points, and addressing them leads to new, \textit{consensus-challenging} insights, such as (1) an early attack, DI, even outperforms all similar follow-up ones, (2) the state-of-the-art (white-box) defense, DiffPure, is even vulnerable to (black-box) transfer attacks, and (3) even under the same $L_p$ constraint, different attacks yield dramatically different stealthiness results regarding diverse imperceptibility metrics, finer-grained measures, and a user study. We hope that our analyses will serve as guidance on properly evaluating transferable adversarial images and advance the design of attacks and defenses. Code is available at https://github.com/ZhengyuZhao/TransferAttackEval.

Dirren van Vlijmen, Alex Kolmus, Zhuoran Liu et al.

We introduce ShortcutGen, a new data poisoning attack that generates sample-dependent, error-minimizing perturbations by learning a generator. The key novelty of ShortcutGen is the use of a randomly-initialized discriminator, which provides spurious shortcuts needed for generating poisons. Different from recent, iterative methods, our ShortcutGen can generate perturbations with only one forward pass in a label-free manner, and compared to the only existing generative method, DeepConfuse, our ShortcutGen is faster and simpler to train while remaining competitive. We also demonstrate that integrating a simple augmentation strategy can further boost the robustness of ShortcutGen against early stopping, and combining augmentation and non-augmentation leads to new state-of-the-art results in terms of final validation accuracy, especially in the challenging, transfer scenario. Lastly, we speculate, through uncovering its working mechanism, that learning a more general representation space could allow ShortcutGen to work for unseen data.

Jingyi Deng, Chenhao Lin, Zhengyu Zhao et al.

Deep generative models have demonstrated impressive performance in various computer vision applications, including image synthesis, video generation, and medical analysis. Despite their significant advancements, these models may be used for malicious purposes, such as misinformation, deception, and copyright violation. In this paper, we provide a systematic and timely review of research efforts on defenses against AI-generated visual media, covering detection, disruption, and authentication. We review existing methods and summarize the mainstream defense-related tasks within a unified passive and proactive framework. Moreover, we survey the derivative tasks concerning the trustworthiness of defenses, such as their robustness and fairness. For each defense strategy, we formulate its general pipeline and propose a multidimensional taxonomy applicable across defense tasks, based on methodological strategies. Additionally, we summarize the commonly used evaluation datasets, criteria, and metrics. Finally, by analyzing the reviewed studies, we provide insights into current research challenges and suggest possible directions for future research.

Qiwei Tian, Chenhao Lin, Zhengyu Zhao et al.

To address the trade-off between robustness and performance for robust VLM, we observe that function words could incur vulnerability of VLMs against cross-modal adversarial attacks, and propose Function-word De-Attention (FDA) accordingly to mitigate the impact of function words. Similar to differential amplifiers, our FDA calculates the original and the function-word cross-attention within attention heads, and differentially subtracts the latter from the former for more aligned and robust VLMs. Comprehensive experiments include 2 SOTA baselines under 6 different attacks on 2 downstream tasks, 3 datasets, and 3 models. Overall, our FDA yields an average 18/13/53% ASR drop with only 0.2/0.3/0.6% performance drops on the 3 tested models on retrieval, and a 90% ASR drop with a 0.3% performance gain on visual grounding. We demonstrate the scalability, generalization, and zero-shot performance of FDA experimentally, as well as in-depth ablation studies and analysis. Code will be made publicly at https://github.com/michaeltian108/FDA.

Zhengyu Zhao, Nga Dang, Martha Larson

Adversarial images are created with the intention of causing an image classifier to produce a misclassification. In this paper, we propose that adversarial images should be evaluated based on semantic mismatch, rather than label mismatch, as used in current work. In other words, we propose that an image of a "mug" would be considered adversarial if classified as "turnip", but not as "cup", as current systems would assume. Our novel idea of taking semantic misclassification into account in the evaluation of adversarial images offers two benefits. First, it is a more realistic conceptualization of what makes an image adversarial, which is important in order to fully understand the implications of adversarial images for security and privacy. Second, it makes it possible to evaluate the transferability of adversarial images to a real-world classifier, without requiring the classifier's label set to have been available during the creation of the images. The paper carries out an evaluation of a transfer attack on a real-world image classifier that is made possible by our semantic misclassification approach. The attack reveals patterns in the semantics of adversarial misclassifications that could not be investigated using conventional label mismatch.

Hamid Bostani, Zhengyu Zhao, Zhuoran Liu et al.

Machine Learning (ML) promises to enhance the efficacy of Android Malware Detection (AMD); however, ML models are vulnerable to realistic evasion attacks--crafting realizable Adversarial Examples (AEs) that satisfy Android malware domain constraints. To eliminate ML vulnerabilities, defenders aim to identify susceptible regions in the feature space where ML models are prone to deception. The primary approach to identifying vulnerable regions involves investigating realizable AEs, but generating these feasible apps poses a challenge. For instance, previous work has relied on generating either feature-space norm-bounded AEs or problem-space realizable AEs in adversarial hardening. The former is efficient but lacks full coverage of vulnerable regions while the latter can uncover these regions by satisfying domain constraints but is known to be time-consuming. To address these limitations, we propose an approach to facilitate the identification of vulnerable regions. Specifically, we introduce a new interpretation of Android domain constraints in the feature space, followed by a novel technique that learns them. Our empirical evaluations across various evasion attacks indicate effective detection of AEs using learned domain constraints, with an average of 89.6%. Furthermore, extensive experiments on different Android malware detectors demonstrate that utilizing our learned domain constraints in Adversarial Training (AT) outperforms other AT-based defenses that rely on norm-bounded AEs or state-of-the-art non-uniform perturbations. Finally, we show that retraining a malware detector with a wide variety of feature-space realizable AEs results in a 77.9% robustness improvement against realizable AEs generated by unknown problem-space transformations, with up to 70x faster training than using problem-space realizable AEs.

Hai Huang, Zhengyu Zhao, Michael Backes et al.

Fine-tuning large pre-trained computer vision models is infeasible for resource-limited users. Visual prompt learning (VPL) has thus emerged to provide an efficient and flexible alternative to model fine-tuning through Visual Prompt as a Service (VPPTaaS). Specifically, the VPPTaaS provider optimizes a visual prompt given downstream data, and downstream users can use this prompt together with the large pre-trained model for prediction. However, this new learning paradigm may also pose security risks when the VPPTaaS provider instead provides a malicious visual prompt. In this paper, we take the first step to explore such risks through the lens of backdoor attacks. Specifically, we propose BadVisualPrompt, a simple yet effective backdoor attack against VPL. For example, poisoning $5\%$ CIFAR10 training data leads to above $99\%$ attack success rates with only negligible model accuracy drop by $1.5\%$. In particular, we identify and then address a new technical challenge related to interactions between the backdoor trigger and visual prompt, which does not exist in conventional, model-level backdoors. Moreover, we provide in-depth analyses of seven backdoor defenses from model, prompt, and input levels. Overall, all these defenses are either ineffective or impractical to mitigate our BadVisualPrompt, implying the critical vulnerability of VPL.

Le Yang, Ziwei Zheng, Boxu Chen et al.

Recent studies have shown that large vision-language models (LVLMs) often suffer from the issue of object hallucinations (OH). To mitigate this issue, we introduce an efficient method that edits the model weights based on an unsafe subspace, which we call HalluSpace in this paper. With truthful and hallucinated text prompts accompanying the visual content as inputs, the HalluSpace can be identified by extracting the hallucinated embedding features and removing the truthful representations in LVLMs. By orthogonalizing the model weights, input features will be projected into the Null space of the HalluSpace to reduce OH, based on which we name our method Nullu. We reveal that HalluSpaces generally contain prior information in the large language models (LLMs) applied to build LVLMs, which have been shown as essential causes of OH in previous studies. Therefore, null space projection suppresses the LLMs' priors to filter out the hallucinated features, resulting in contextually accurate outputs. Experiments show that our method can effectively mitigate OH across different LVLM families without extra inference costs and also show strong performance in general LVLM benchmarks. Code is released at https://github.com/Ziwei-Zheng/Nullu.

Hamid Bostani, Zhengyu Zhao, Veelasha Moonsamy

Machine learning (ML) has demonstrated significant advancements in Android malware detection (AMD); however, the resilience of ML against realistic evasion attacks remains a major obstacle for AMD. One of the primary factors contributing to this challenge is the scarcity of reliable generalizations. Malware classifiers with limited generalizability tend to overfit spurious correlations derived from biased features. Consequently, adversarial examples (AEs), generated by evasion attacks, can modify these features to evade detection. In this study, we propose a domain adaptation technique to improve the generalizability of AMD by aligning the distribution of malware samples and AEs. Specifically, we utilize meaningful feature dependencies, reflecting domain constraints in the feature space, to establish a robust feature space. Training on the proposed robust feature space enables malware classifiers to learn from predefined patterns associated with app functionality rather than from individual features. This approach helps mitigate spurious correlations inherent in the initial feature space. Our experiments conducted on DREBIN, a renowned Android malware detector, demonstrate that our approach surpasses the state-of-the-art defense, Sec-SVM, when facing realistic evasion attacks. In particular, our defense can improve adversarial robustness by up to 55% against realistic evasion attacks compared to Sec-SVM.

Songlong Xing, Zhengyu Zhao, Nicu Sebe

Despite its prevalent use in image-text matching tasks in a zero-shot manner, CLIP has been shown to be highly vulnerable to adversarial perturbations added onto images. Recent studies propose to finetune the vision encoder of CLIP with adversarial samples generated on the fly, and show improved robustness against adversarial attacks on a spectrum of downstream datasets, a property termed as zero-shot robustness. In this paper, we show that malicious perturbations that seek to maximise the classification loss lead to `falsely stable' images, and propose to leverage the pre-trained vision encoder of CLIP to counterattack such adversarial images during inference to achieve robustness. Our paradigm is simple and training-free, providing the first method to defend CLIP from adversarial attacks at test time, which is orthogonal to existing methods aiming to boost zero-shot adversarial robustness of CLIP. We conduct experiments across 16 classification datasets, and demonstrate stable and consistent gains compared to test-time defence methods adapted from existing adversarial robustness studies that do not rely on external networks, without noticeably impairing performance on clean images. We also show that our paradigm can be employed on CLIP models that have been adversarially finetuned to further enhance their robustness at test time. Our code is available \href{https://github.com/Sxing2/CLIP-Test-time-Counterattacks}{here}.

Yuchen Ren, Zhengyu Zhao, Chenhao Lin et al.

Transferable adversarial examples are known to cause threats in practical, black-box attack scenarios. A notable approach to improving transferability is using integrated gradients (IG), originally developed for model interpretability. In this paper, we find that existing IG-based attacks have limited transferability due to their naive adoption of IG in model interpretability. To address this limitation, we focus on the IG integration path and refine it in three aspects: multiplicity, monotonicity, and diversity, supported by theoretical analyses. We propose the Multiple Monotonic Diversified Integrated Gradients (MuMoDIG) attack, which can generate highly transferable adversarial examples on different CNN and ViT models and defenses. Experiments validate that MuMoDIG outperforms the latest IG-based attack by up to 37.3\% and other state-of-the-art attacks by 8.4\%. In general, our study reveals that migrating established techniques to improve transferability may require non-trivial efforts. Code is available at \url{https://github.com/RYC-98/MuMoDIG}.

Yuchen Ren, Zhengyu Zhao, Chenhao Lin et al.

Vision Transformers (ViTs) have been widely applied in various computer vision and vision-language tasks. To gain insights into their robustness in practical scenarios, transferable adversarial examples on ViTs have been extensively studied. A typical approach to improving adversarial transferability is by refining the surrogate model. However, existing work on ViTs has restricted their surrogate refinement to backward propagation. In this work, we instead focus on Forward Propagation Refinement (FPR) and specifically refine two key modules of ViTs: attention maps and token embeddings. For attention maps, we propose Attention Map Diversification (AMD), which diversifies certain attention maps and also implicitly imposes beneficial gradient vanishing during backward propagation. For token embeddings, we propose Momentum Token Embedding (MTE), which accumulates historical token embeddings to stabilize the forward updates in both the Attention and MLP blocks. We conduct extensive experiments with adversarial examples transferred from ViTs to various CNNs and ViTs, demonstrating that our FPR outperforms the current best (backward) surrogate refinement by up to 7.0\% on average. We also validate its superiority against popular defenses and its compatibility with other transfer methods. Codes and appendix are available at https://github.com/RYC-98/FPR.

Weijie Wang, Songlong Xing, Zhengyu Zhao et al.

Poisoning input views of 3D reconstruction systems has been recently studied. However, we identify that existing studies simply backpropagate adversarial gradients through the 3D reconstruction pipeline as a whole, without uncovering the new vulnerability rooted in specific modules of the 3D reconstruction pipeline. In this paper, we argue that the structure-from-motion (SfM) initialization, as the geometric core of many widely used reconstruction systems, can be targeted to achieve transferable poisoning effects across diverse 3D reconstruction systems. To this end, we propose PoInit-of-View, which optimizes adversarial perturbations to intentionally introduce cross-view gradient inconsistencies at projections of corresponding 3D points. These inconsistencies disrupt keypoint detection and feature matching, thereby corrupting pose estimation and triangulation within SfM, eventually resulting in low-quality rendered views. We also provide a theoretical analysis that connects cross-view inconsistency to correspondence collapse. Experimental results demonstrate the effectiveness of our PoInit-of-View on diverse 3D reconstruction systems and datasets, surpassing the single-view baseline by 25.1% in PSNR and 16.5% in SSIM in black-box transfer settings, such as 3DGS to NeRF.

Boxu Chen, Ziwei Zheng, Le Yang et al.

Large Vision-Language Models (LVLMs) have achieved remarkable success but continue to struggle with object hallucination (OH), generating outputs inconsistent with visual inputs. While previous work has proposed methods to reduce OH, the visual decision-making mechanisms that lead to hallucinations remain poorly understood. In this paper, we propose VaLSe, a Vision-aware Latent Steering framework that adopts an interpretation-then-mitigation strategy to address OH in LVLMs. By tackling dual challenges of modeling complex vision-language interactions and eliminating spurious activation artifacts, VaLSe can generate visual contribution maps that trace how specific visual inputs influence individual output tokens. These maps reveal the model's vision-aware focus regions, which are then used to perform latent space steering, realigning internal representations toward semantically relevant content and reducing hallucinated outputs. Extensive experiments demonstrate that VaLSe is a powerful interpretability tool and an effective method for enhancing model robustness against OH across multiple benchmarks. Furthermore, our analysis uncovers limitations in existing OH evaluation metrics, underscoring the need for more nuanced, interpretable, and visually grounded OH benchmarks in future work. Code is available at: https://github.com/Ziwei-Zheng/VaLSe.

Qiwei Tian, Chenhao Lin, Zhengyu Zhao et al.

Adversarial training has achieved substantial performance in defending image retrieval against adversarial examples. However, existing studies in deep metric learning (DML) still suffer from two major limitations: weak adversary and model collapse. In this paper, we address these two limitations by proposing Collapse-Aware TRIplet DEcoupling (CA-TRIDE). Specifically, TRIDE yields a stronger adversary by spatially decoupling the perturbation targets into the anchor and the other candidates. Furthermore, CA prevents the consequential model collapse, based on a novel metric, collapseness, which is incorporated into the optimization of perturbation. We also identify two drawbacks of the existing robustness metric in image retrieval and propose a new metric for a more reasonable robustness evaluation. Extensive experiments on three datasets demonstrate that CA-TRIDE outperforms existing defense methods in both conventional and new metrics. Codes are available at https://github.com/michaeltian108/CA-TRIDE.

Junhao Zheng, Jiahao Sun, Chenhao Lin et al.

Developing reliable defenses against patch attacks on object detectors has attracted increasing interest. However, we identify that existing defense evaluations lack a unified and comprehensive framework, resulting in inconsistent and incomplete assessments of current methods. To address this issue, we revisit 11 representative defenses and present the first patch defense benchmark, involving 2 attack goals, 13 patch attacks, 11 object detectors, and 4 diverse metrics. This leads to the large-scale adversarial patch dataset with 94 types of patches and 94,000 images. Our comprehensive analyses reveal new insights: (1) The difficulty in defending against naturalistic patches lies in the data distribution, rather than the commonly believed high frequencies. Our new dataset with diverse patch distributions can be used to improve existing defenses by 15.09% AP@0.5. (2) The average precision of the attacked object, rather than the commonly pursued patch detection accuracy, shows high consistency with defense performance. (3) Adaptive attacks can substantially bypass existing defenses, and defenses with complex/stochastic models or universal patch properties are relatively robust. We hope that our analyses will serve as guidance on properly evaluating patch attacks/defenses and advancing their design. Code and dataset are available at https://github.com/Gandolfczjh/APDE, where we will keep integrating new attacks/defenses.

Qiwei Tian, Chenhao Lin, Zhengyu Zhao et al.

Thanks to the development of cross-modal models, text-to-video retrieval (T2VR) is advancing rapidly, but its robustness remains largely unexamined. Existing attacks against T2VR are designed to push videos away from queries, i.e., suppressing the ranks of videos, while the attacks that pull videos towards selected queries, i.e., promoting the ranks of videos, remain largely unexplored. These attacks can be more impactful as attackers may gain more views/clicks for financial benefits and widespread (mis)information. To this end, we pioneer the first attack against T2VR to promote videos adversarially, dubbed the Video Promotion attack (ViPro). We further propose Modal Refinement (MoRe) to capture the finer-grained, intricate interaction between visual and textual modalities to enhance black-box transferability. Comprehensive experiments cover 2 existing baselines, 3 leading T2VR models, 3 prevailing datasets with over 10k videos, evaluated under 3 scenarios. All experiments are conducted in a multi-target setting to reflect realistic scenarios where attackers seek to promote the video regarding multiple queries simultaneously. We also evaluated our attacks for defences and imperceptibility. Overall, ViPro surpasses other baselines by over $30/10/4\%$ for white/grey/black-box settings on average. Our work highlights an overlooked vulnerability, provides a qualitative analysis on the upper/lower bound of our attacks, and offers insights into potential counterplays. Code will be publicly available at https://github.com/michaeltian108/ViPro.

Chende Zheng, Ruiqi suo, Chenhao Lin et al.

The evolution of video generation techniques, such as Sora, has made it increasingly easy to produce high-fidelity AI-generated videos, raising public concern over the dissemination of synthetic content. However, existing detection methodologies remain limited by their insufficient exploration of temporal artifacts in synthetic videos. To bridge this gap, we establish a theoretical framework through second-order dynamical analysis under Newtonian mechanics, subsequently extending the Second-order Central Difference features tailored for temporal artifact detection. Building on this theoretical foundation, we reveal a fundamental divergence in second-order feature distributions between real and AI-generated videos. Concretely, we propose Detection by Difference of Differences (D3), a novel training-free detection method that leverages the above second-order temporal discrepancies. We validate the superiority of our D3 on 4 open-source datasets (Gen-Video, VideoPhy, EvalCrafter, VidProM), 40 subsets in total. For example, on GenVideo, D3 outperforms the previous best method by 10.39% (absolute) mean Average Precision. Additional experiments on time cost and post-processing operations demonstrate D3's exceptional computational efficiency and strong robust performance. Our code is available at https://github.com/Zig-HS/D3.

Song Yan, Hui Wei, Jinlong Fei et al.

Various (text) prompt filters and (image) safety checkers have been implemented to mitigate the misuse of Text-to-Image (T2I) models in creating Not-Safe-For-Work (NSFW) content. In order to expose potential security vulnerabilities of such safeguards, multimodal jailbreaks have been studied. However, existing jailbreaks are limited to prompt-specific and image-specific perturbations, which suffer from poor scalability and time-consuming optimization. To address these limitations, we propose Universally Unfiltered and Unseen (U3)-Attack, a multimodal jailbreak attack method against T2I safeguards. Specifically, U3-Attack optimizes an adversarial patch on the image background to universally bypass safety checkers and optimizes a safe paraphrase set from a sensitive word to universally bypass prompt filters while eliminating redundant computations. Extensive experimental results demonstrate the superiority of our U3-Attack on both open-source and commercial T2I models. For example, on the commercial Runway-inpainting model with both prompt filter and safety checker, our U3-Attack achieves $~4\times$ higher success rates than the state-of-the-art multimodal jailbreak attack, MMA-Diffusion.

Weijie Wang, Zhengyu Zhao, Nicu Sebe et al.

Malicious use of deepfakes leads to serious public concerns and reduces people's trust in digital media. Although effective deepfake detectors have been proposed, they are substantially vulnerable to adversarial attacks. To evaluate the detector's robustness, recent studies have explored various attacks. However, all existing attacks are limited to 2D image perturbations, which are hard to translate into real-world facial changes. In this paper, we propose adversarial head turn (AdvHeat), the first attempt at 3D adversarial face views against deepfake detectors, based on face view synthesis from a single-view fake image. Extensive experiments validate the vulnerability of various detectors to AdvHeat in realistic, black-box scenarios. For example, AdvHeat based on a simple random search yields a high attack success rate of 96.8% with 360 searching steps. When additional query access is allowed, we can further reduce the step budget to 50. Additional analyses demonstrate that AdvHeat is better than conventional attacks on both the cross-detector transferability and robustness to defenses. The adversarial images generated by AdvHeat are also shown to have natural looks. Our code, including that for generating a multi-view dataset consisting of 360 synthetic views for each of 1000 IDs from FaceForensics++, is available at https://github.com/twowwj/AdvHeaT.

Zhengyu Zhao, Zhuoran Liu, Martha Larson

Achieving transferability of targeted attacks is reputed to be remarkably difficult. Currently, state-of-the-art approaches are resource-intensive because they necessitate training model(s) for each target class with additional data. In our investigation, we find, however, that simple transferable attacks which require neither additional data nor model training can achieve surprisingly high targeted transferability. This insight has been overlooked until now, mainly due to the widespread practice of unreasonably restricting attack optimization to a limited number of iterations. In particular, we, for the first time, identify that a simple logit loss can yield competitive results with the state of the arts. Our analysis spans a variety of transfer settings, especially including three new, realistic settings: an ensemble transfer setting with little model similarity, a worse-case setting with low-ranked target classes, and also a real-world attack against the Google Cloud Vision API. Results in these new settings demonstrate that the commonly adopted, easy settings cannot fully reveal the actual properties of different attacks and may cause misleading comparisons. We also show the usefulness of the simple logit loss for generating targeted universal adversarial perturbations in a data-free and training-free manner. Overall, the aim of our analysis is to inspire a more meaningful evaluation on targeted transferability. Code is available at https://github.com/ZhengyuZhao/Targeted-Tansfer

Zhengyu Zhao, Zhuoran Liu, Martha Larson

We introduce an approach that enhances images using a color filter in order to create adversarial effects, which fool neural networks into misclassification. Our approach, Adversarial Color Enhancement (ACE), generates unrestricted adversarial images by optimizing the color filter via gradient descent. The novelty of ACE is its incorporation of established practice for image enhancement in a transparent manner. Experimental results validate the white-box adversarial strength and black-box transferability of ACE. A range of examples demonstrates the perceptual quality of images that ACE produces. ACE makes an important contribution to recent work that moves beyond $L_p$ imperceptibility and focuses on unrestricted adversarial modifications that yield large perceptible perturbations, but remain non-suspicious, to the human eye. The future potential of filter-based adversaries is also explored in two directions: guiding ACE with common enhancement practices (e.g., Instagram filters) towards specific attractive image styles and adapting ACE to image semantics. Code is available at https://github.com/ZhengyuZhao/ACE.

Chaoshuo Zhang, Yibo Liang, Mengke Tian et al.

Despite the remarkable synthesis capabilities of text-to-image (T2I) models, safeguarding them against content violations remains a persistent challenge. Existing safety alignments primarily focus on explicit malicious concepts, often overlooking the subtle yet critical risks of compositional semantics. To address this oversight, we identify and formalize a novel vulnerability: Multi-Concept Compositional Unsafety (MCCU), where unsafe semantics stem from the implicit associations of individually benign concepts. Based on this formulation, we introduce TwoHamsters, a comprehensive benchmark comprising 17.5k prompts curated to probe MCCU vulnerabilities. Through a rigorous evaluation of 10 state-of-the-art models and 16 defense mechanisms, our analysis yields 8 pivotal insights. In particular, we demonstrate that current T2I models and defense mechanisms face severe MCCU risks: on TwoHamsters, FLUX achieves an MCCU generation success rate of 99.52%, while LLaVA-Guard only attains a recall of 41.06%, highlighting a critical limitation of the current paradigm for managing hazardous compositional generation.

Junhao Zheng, Chenhao Lin, Jiahao Sun et al.

Deep learning-based monocular depth estimation (MDE), extensively applied in autonomous driving, is known to be vulnerable to adversarial attacks. Previous physical attacks against MDE models rely on 2D adversarial patches, so they only affect a small, localized region in the MDE map but fail under various viewpoints. To address these limitations, we propose 3D Depth Fool (3D$^2$Fool), the first 3D texture-based adversarial attack against MDE models. 3D$^2$Fool is specifically optimized to generate 3D adversarial textures agnostic to model types of vehicles and to have improved robustness in bad weather conditions, such as rain and fog. Experimental results validate the superior performance of our 3D$^2$Fool across various scenarios, including vehicles, MDE models, weather conditions, and viewpoints. Real-world experiments with printed 3D textures on physical vehicle models further demonstrate that our 3D$^2$Fool can cause an MDE error of over 10 meters.

Xiaomeng Wang, Martha Larson, Zhengyu Zhao

When the visual style of text is considered, a wide variety can be observed in font, color, and size. However, when a word is read, its meaning is independent of the style in which it has been written or rendered. In this paper, we investigate whether, and how, the style in which a word is visualized in an image impacts the description that a Large Visual Language Model (LVLM) provides for the concept to which that word refers. Specifically, we investigate how functional text styles (readability-oriented, e.g., black sans-serif) versus decorative styles (display-oriented, e.g., colored cursive/script) affect LVLMs' descriptions of a concept in terms of the attributes of that concept. Our experiments study the situation in which the LVLM is able to correctly identify the concept referred to by a visual text, i.e., by a word or words rendered as an image, and in which the visual text style should not influence the attribute-based description that the LVLM produces. Our experimental results reveal that even when the concept is correctly identified, text style influences the model's attribute-based descriptions of the concept. Our findings demonstrate non-trivial style leakage from text style into semantic inference and motivate style-aware evaluation and mitigation for LVLM-based multimedia systems.

Bo Yang, Hengwei Zhang, Jindong Wang et al.

Transferable adversarial examples cause practical security risks since they can mislead a target model without knowing its internal knowledge. A conventional recipe for maximizing transferability is to keep only the optimal adversarial example from all those obtained in the optimization pipeline. In this paper, for the first time, we revisit this convention and demonstrate that those discarded, sub-optimal adversarial examples can be reused to boost transferability. Specifically, we propose ``Adversarial Example Soups'' (AES), with AES-tune for averaging discarded adversarial examples in hyperparameter tuning and AES-rand for stability testing. In addition, our AES is inspired by ``model soups'', which averages weights of multiple fine-tuned models for improved accuracy without increasing inference time. Extensive experiments validate the global effectiveness of our AES, boosting 10 state-of-the-art transfer attacks and their combinations by up to 13\% against 10 diverse (defensive) target models. We also show the possibility of generalizing AES to other types, \textit{e.g.}, directly averaging multiple in-the-wild adversarial examples that yield comparable success. A promising byproduct of AES is the improved stealthiness of adversarial examples since the perturbation variances are naturally reduced.

Chaoshuo Zhang, Chenhao Lin, Zhengyu Zhao et al.

Text-to-image diffusion models (T2I DMs), represented by Stable Diffusion, which generate highly realistic images based on textual input, have been widely used, but their flexibility also makes them prone to misuse for producing harmful or unsafe content. Concept unlearning has been used to prevent text-to-image diffusion models from being misused to generate undesirable visual content. However, existing methods struggle to trade off unlearning effectiveness with the preservation of generation quality. To address this limitation, we propose Key Step Concept Unlearning (KSCU), which selectively fine-tunes the model at key steps to the target concept. KSCU is inspired by the fact that different diffusion denoising steps contribute unequally to the final generation. Compared to previous approaches, which treat all denoising steps uniformly, KSCU avoids over-optimization of unnecessary steps for higher effectiveness and reduces the number of parameter updates for higher efficiency. For example, on the I2P dataset, KSCU outperforms ESD by 8.3% in nudity unlearning accuracy while improving FID by 8.4%, and achieves a high overall score of 0.92, substantially surpassing all other SOTA methods.

Zhizhen Chen, Zhengyu Zhao, Subrat Kishore Dutta et al.

Targeted data poisoning (TDP) aims to compromise the model's prediction on a specific (test) target by perturbing a small subset of training data. Existing work on TDP has focused on an overly ideal threat model in which the same image sample of the target is used during both poisoning and inference stages. However, in the real world, a target object often appears in complex variations due to changes of physical settings such as viewpoint, background, and lighting conditions. In this work, we take the first step toward understanding the real-world threats of TDP by studying its generalizability across varying physical conditions. In particular, we observe that solely optimizing gradient directions, as adopted by the best previous TDP method, achieves limited generalization. To address this limitation, we propose optimizing both the gradient direction and magnitude for more generalizable gradient matching, thereby leading to higher poisoning success rates. For instance, our method outperforms the state of the art by 19.49% when poisoning CIFAR-10 images targeting multi-view cars.

Ruixuan Zhang, He Wang, Zhengyu Zhao et al.

Rapid advances in Artificial Intelligence Generated Images (AIGI) have facilitated malicious use, such as forgery and misinformation. Therefore, numerous methods have been proposed to detect fake images. Although such detectors have been proven to be universally vulnerable to adversarial attacks, defenses in this field are scarce. In this paper, we first identify that adversarial training (AT), widely regarded as the most effective defense, suffers from performance collapse in AIGI detection. Through an information-theoretic lens, we further attribute the cause of collapse to feature entanglement, which disrupts the preservation of feature-label mutual information. Instead, standard detectors show clear feature separation. Motivated by this difference, we propose Training-free Robust Detection via Information-theoretic Measures (TRIM), the first training-free adversarial defense for AIGI detection. TRIM builds on standard detectors and quantifies feature shifts using prediction entropy and KL divergence. Extensive experiments across multiple datasets and attacks validate the superiority of our TRIM, e.g., outperforming the state-of-the-art defense by 33.88% (28.91%) on ProGAN (GenImage), while well maintaining original accuracy.

Bo Yang, Hengwei Zhang, Jindong Wang et al.

In surrogate ensemble attacks, using more surrogate models yields higher transferability but lower resource efficiency. This practical trade-off between transferability and efficiency has largely limited existing attacks despite many pre-trained models are easily accessible online. In this paper, we argue that such a trade-off is caused by an unnecessary common assumption, i.e., all models should be \textit{identical} across iterations. By lifting this assumption, we can use as many surrogates as we want to unleash transferability without sacrificing efficiency. Concretely, we propose Selective Ensemble Attack (SEA), which dynamically selects diverse models (from easily accessible pre-trained models) across iterations based on our new interpretation of decoupling within-iteration and cross-iteration model diversity. In this way, the number of within-iteration models is fixed for maintaining efficiency, while only cross-iteration model diversity is increased for higher transferability. Experiments on ImageNet demonstrate the superiority of SEA in various scenarios. For example, when dynamically selecting 4 from 20 accessible models, SEA yields 8.5% higher transferability than existing attacks under the same efficiency. The superiority of SEA also generalizes to real-world systems, such as commercial vision APIs and large vision-language models. Overall, SEA opens up the possibility of adaptively balancing transferability and efficiency according to specific resource requirements.

Chenhao Lin, Chenyang Zhao, Shiwei Wang et al.

Backdoor attacks typically place a specific trigger on certain training data, such that the model makes prediction errors on inputs with that trigger during inference. Despite the core role of the trigger, existing studies have commonly believed a perfect match between training-inference triggers is optimal. In this paper, for the first time, we systematically explore the training-inference trigger relation, particularly focusing on their mismatch, based on a Training-Inference Trigger Intensity Manipulation (TITIM) workflow. TITIM specifically investigates the training-inference trigger intensity, such as the size or the opacity of a trigger, and reveals new insights into trigger generalization and overfitting. These new insights challenge the above common belief by demonstrating that the training-inference trigger mismatch can facilitate attacks in two practical scenarios, posing more significant security threats than previously thought. First, when the inference trigger is fixed, using training triggers with mixed intensities leads to stronger attacks than using any single intensity. For example, on CIFAR-10 with ResNet-18, mixing training triggers with 1.0 and 0.1 opacities improves the worst-case attack success rate (ASR) (over different testing opacities) of the best single-opacity attack from 10.61\% to 92.77\%. Second, intentionally using certain mismatched training-inference triggers can improve the attack stealthiness, i.e., better bypassing defenses. For example, compared to the training/inference intensity of 1.0/1.0, using 1.0/0.7 decreases the area under the curve (AUC) of the Scale-Up defense from 0.96 to 0.62, while maintaining a high attack ASR (99.65\% vs. 91.62\%). The above new insights are validated to be generalizable across different backdoor attacks, models, datasets, tasks, and (digital/physical) domains.

Chen Ma, Ningfei Wang, Zhengyu Zhao et al.

Recent research in adversarial machine learning has focused on visual perception in Autonomous Driving (AD) and has shown that printed adversarial patches can attack object detectors. However, it is important to note that AD visual perception encompasses more than just object detection; it also includes Multiple Object Tracking (MOT). MOT enhances the robustness by compensating for object detection errors and requiring consistent object detection results across multiple frames before influencing tracking results and driving decisions. Thus, MOT makes attacks on object detection alone less effective. To attack such robust AD visual perception, a digital hijacking attack has been proposed to cause dangerous driving scenarios. However, this attack has limited effectiveness. In this paper, we introduce a novel physical-world adversarial patch attack, ControlLoc, designed to exploit hijacking vulnerabilities in entire AD visual perception. ControlLoc utilizes a two-stage process: initially identifying the optimal location for the adversarial patch, and subsequently generating the patch that can modify the perceived location and shape of objects with the optimal location. Extensive evaluations demonstrate the superior performance of ControlLoc, achieving an impressive average attack success rate of around 98.1% across various AD visual perceptions and datasets, which is four times greater effectiveness than the existing hijacking attack. The effectiveness of ControlLoc is further validated in physical-world conditions, including real vehicle tests under different conditions such as outdoor light conditions with an average attack success rate of 77.5%. AD system-level impact assessments are also included, such as vehicle collision, using industry-grade AD systems and production-grade AD simulators with an average vehicle collision rate and unnecessary emergency stop rate of 81.3%.

Chen Ma, Ningfei Wang, Zhengyu Zhao et al.

Autonomous Driving (AD) systems critically depend on visual perception for real-time object detection and multiple object tracking (MOT) to ensure safe driving. However, high latency in these visual perception components can lead to significant safety risks, such as vehicle collisions. While previous research has extensively explored latency attacks within the digital realm, translating these methods effectively to the physical world presents challenges. For instance, existing attacks rely on perturbations that are unrealistic or impractical for AD, such as adversarial perturbations affecting areas like the sky, or requiring large patches that obscure most of a camera's view, thus making them impossible to be conducted effectively in the real world. In this paper, we introduce SlowPerception, the first physical-world latency attack against AD perception, via generating projector-based universal perturbations. SlowPerception strategically creates numerous phantom objects on various surfaces in the environment, significantly increasing the computational load of Non-Maximum Suppression (NMS) and MOT, thereby inducing substantial latency. Our SlowPerception achieves second-level latency in physical-world settings, with an average latency of 2.5 seconds across different AD perception systems, scenarios, and hardware configurations. This performance significantly outperforms existing state-of-the-art latency attacks. Additionally, we conduct AD system-level impact assessments, such as vehicle collisions, using industry-grade AD systems with production-grade AD simulators with a 97% average rate. We hope that our analyses can inspire further research in this critical domain, enhancing the robustness of AD systems against emerging vulnerabilities.

Zhuoran Liu, Zhengyu Zhao, Alex Kolmus et al.

Recent work has shown that imperceptible perturbations can be applied to craft unlearnable examples (ULEs), i.e. images whose content cannot be used to improve a classifier during training. In this paper, we reveal the road that researchers should follow for understanding ULEs and improving ULEs as they were originally formulated (ULEOs). The paper makes four contributions. First, we show that ULEOs exploit color and, consequently, their effects can be mitigated by simple grayscale pre-filtering, without resorting to adversarial training. Second, we propose an extension to ULEOs, which is called ULEO-GrayAugs, that forces the generated ULEs away from channel-wise color perturbations by making use of grayscale knowledge and data augmentations during optimization. Third, we show that ULEOs generated using Multi-Layer Perceptrons (MLPs) are effective in the case of complex Convolutional Neural Network (CNN) classifiers, suggesting that CNNs suffer specific vulnerability to ULEs. Fourth, we demonstrate that when a classifier is trained on ULEOs, adversarial training will prevent a drop in accuracy measured both on clean images and on adversarial images. Taken together, our contributions represent a substantial advance in the state of art of unlearnable examples, but also reveal important characteristics of their behavior that must be better understood in order to achieve further improvements.

Zhuoran Liu, Niels Samwel, Léo Weissbart et al.

We introduce screen gleaning, a TEMPEST attack in which the screen of a mobile device is read without a visual line of sight, revealing sensitive information displayed on the phone screen. The screen gleaning attack uses an antenna and a software-defined radio (SDR) to pick up the electromagnetic signal that the device sends to the screen to display, e.g., a message with a security code. This special equipment makes it possible to recreate the signal as a gray-scale image, which we refer to as an emage. Here, we show that it can be used to read a security code. The screen gleaning attack is challenging because it is often impossible for a human viewer to interpret the emage directly. We show that this challenge can be addressed with machine learning, specifically, a deep learning classifier. Screen gleaning will become increasingly serious as SDRs and deep learning continue to rapidly advance. In this paper, we demonstrate the security code attack and we propose a testbed that provides a standard setup in which screen gleaning could be tested with different attacker models. Finally, we analyze the dimensions of screen gleaning attacker models and discuss possible countermeasures with the potential to address them.

Zhengyu Zhao, Zhuoran Liu, Martha Larson

Deep Neural Networks have been shown to be vulnerable to adversarial images. Conventional attacks strive for indistinguishable adversarial images with strictly restricted perturbations. Recently, researchers have moved to explore distinguishable yet non-suspicious adversarial images and demonstrated that color transformation attacks are effective. In this work, we propose Adversarial Color Filter (AdvCF), a novel color transformation attack that is optimized with gradient information in the parameter space of a simple color filter. In particular, our color filter space is explicitly specified so that we are able to provide a systematic analysis of model robustness against adversarial color transformations, from both the attack and defense perspectives. In contrast, existing color transformation attacks do not offer the opportunity for systematic analysis due to the lack of such an explicit space. We further demonstrate the effectiveness of our AdvCF in fooling image classifiers and also compare it with other color transformation attacks regarding their robustness to defenses and image acceptability through an extensive user study. We also highlight the human-interpretability of AdvCF and show its superiority over the state-of-the-art human-interpretable color transformation attack on both image acceptability and efficiency. Additional results provide interesting new insights into model robustness against AdvCF in another three visual tasks.

Haoyu Song, Yan Wang, Wei-Nan Zhang et al.

Maintaining a consistent attribute profile is crucial for dialogue agents to naturally converse with humans. Existing studies on improving attribute consistency mainly explored how to incorporate attribute information in the responses, but few efforts have been made to identify the consistency relations between response and attribute profile. To facilitate the study of profile consistency identification, we create a large-scale human-annotated dataset with over 110K single-turn conversations and their key-value attribute profiles. Explicit relation between response and profile is manually labeled. We also propose a key-value structure information enriched BERT model to identify the profile consistency, and it gained improvements over strong baselines. Further evaluations on downstream tasks demonstrate that the profile consistency identification model is conducive for improving dialogue consistency.

Zhengyu Zhao, Zhuoran Liu, Martha Larson

The success of image perturbations that are designed to fool image classifier is assessed in terms of both adversarial effect and visual imperceptibility. The conventional assumption on imperceptibility is that perturbations should strive for tight $L_p$-norm bounds in RGB space. In this work, we drop this assumption by pursuing an approach that exploits human color perception, and more specifically, minimizing perturbation size with respect to perceptual color distance. Our first approach, Perceptual Color distance C&W (PerC-C&W), extends the widely-used C&W approach and produces larger RGB perturbations. PerC-C&W is able to maintain adversarial strength, while contributing to imperceptibility. Our second approach, Perceptual Color distance Alternating Loss (PerC-AL), achieves the same outcome, but does so more efficiently by alternating between the classification loss and perceptual color difference when updating perturbations. Experimental evaluation shows PerC approaches outperform conventional $L_p$ approaches in terms of robustness and transferability, and also demonstrates that the PerC distance can provide added value on top of existing structure-based methods to creating image perturbations.

Zhuoran Liu, Zhengyu Zhao, Martha Larson

An adversarial query is an image that has been modified to disrupt content-based image retrieval (CBIR) while appearing nearly untouched to the human eye. This paper presents an analysis of adversarial queries for CBIR based on neural, local, and global features. We introduce an innovative neural image perturbation approach, called Perturbations for Image Retrieval Error (PIRE), that is capable of blocking neural-feature-based CBIR. PIRE differs significantly from existing approaches that create images adversarial with respect to CNN classifiers because it is unsupervised, i.e., it needs no labelled data from the data set to which it is applied. Our experimental analysis demonstrates the surprising effectiveness of PIRE in blocking CBIR, and also covers aspects of PIRE that must be taken into account in practical settings, including saving images, image quality and leaking adversarial queries into the background collection. Our experiments also compare PIRE (a neural approach) with existing keypoint removal and injection approaches (which modify local features). Finally, we discuss the challenges that face multimedia researchers in the future study of adversarial queries.

Zhengyu Zhao, Martha Larson

As deep learning approaches to scene recognition emerge, they have continued to leverage discriminative regions at multiple scales, building on practices established by conventional image classification research. However, approaches remain largely generic, and do not carefully consider the special properties of scenes. In this paper, inspired by the intuitive differences between scenes and objects, we propose Adi-Red, an adaptive approach to discriminative region discovery for scene recognition. Adi-Red uses a CNN classifier, which was pre-trained using only image-level scene labels, to discover discriminative image regions directly. These regions are then used as a source of features to perform scene recognition. The use of the CNN classifier makes it possible to adapt the number of discriminative regions per image using a simple, yet elegant, threshold, at relatively low computational cost. Experimental results on the scene recognition benchmark dataset SUN397 demonstrate the ability of Adi-Red to outperform the state of the art. Additional experimental analysis on the Places dataset reveals the advantages of Adi-Red, and highlight how they are specific to scenes. We attribute the effectiveness of Adi-Red to the ability of adaptive region discovery to avoid introducing noise, while also not missing out on important information.