Zhongjie Ba

Chao Shuai, Jieming Zhong, Shuang Wu et al.

Deepfake has taken the world by storm, triggering a trust crisis. Current deepfake detection methods are typically inadequate in generalizability, with a tendency to overfit to image contents such as the background, which are frequently occurring but relatively unimportant in the training dataset. Furthermore, current methods heavily rely on a few dominant forgery regions and may ignore other equally important regions, leading to inadequate uncovering of forgery cues. In this paper, we strive to address these shortcomings from three aspects: (1) We propose an innovative two-stream network that effectively enlarges the potential regions from which the model extracts forgery evidence. (2) We devise three functional modules to handle the multi-stream and multi-scale features in a collaborative learning scheme. (3) Confronted with the challenge of obtaining forgery annotations, we propose a Semi-supervised Patch Similarity Learning strategy to estimate patch-level forged location annotations. Empirically, our method demonstrates significantly improved robustness and generalizability, outperforming previous methods on six benchmarks, and improving the frame-level AUC on Deepfake Detection Challenge preview dataset from 0.797 to 0.835 and video-level AUC on CelebDF$\_$v1 dataset from 0.811 to 0.847. Our implementation is available at https://github.com/sccsok/Locate-and-Verify.

Kun Pan, Yin Yifang, Yao Wei et al.

The malicious use and widespread dissemination of deepfake pose a significant crisis of trust. Current deepfake detection models can generally recognize forgery images by training on a large dataset. However, the accuracy of detection models degrades significantly on images generated by new deepfake methods due to the difference in data distribution. To tackle this issue, we present a novel incremental learning framework that improves the generalization of deepfake detection models by continual learning from a small number of new samples. To cope with different data distributions, we propose to learn a domain-invariant representation based on supervised contrastive learning, preventing overfit to the insufficient new data. To mitigate catastrophic forgetting, we regularize our model in both feature-level and label-level based on a multi-perspective knowledge distillation approach. Finally, we propose to select both central and hard representative samples to update the replay set, which is beneficial for both domain-invariant representation learning and rehearsal-based knowledge preserving. We conduct extensive experiments on four benchmark datasets, obtaining the new state-of-the-art average forgetting rate of 7.01 and average accuracy of 85.49 on FF++, DFDC-P, DFD, and CDF2. Our code is released at https://github.com/DeepFakeIL/DFIL.

Xinyu Zhang, Qingyu Liu, Zhongjie Ba et al.

Federated Learning (FL) is a promising distributed learning approach that enables multiple clients to collaboratively train a shared global model. However, recent studies show that FL is vulnerable to various poisoning attacks, which can degrade the performance of global models or introduce backdoors into them. In this paper, we first conduct a comprehensive study on prior FL attacks and detection methods. The results show that all existing detection methods are only effective against limited and specific attacks. Most detection methods suffer from high false positives, which lead to significant performance degradation, especially in not independent and identically distributed (non-IID) settings. To address these issues, we propose FLTracer, the first FL attack provenance framework to accurately detect various attacks and trace the attack time, objective, type, and poisoned location of updates. Different from existing methodologies that rely solely on cross-client anomaly detection, we propose a Kalman filter-based cross-round detection to identify adversaries by seeking the behavior changes before and after the attack. Thus, this makes it resilient to data heterogeneity and is effective even in non-IID settings. To further improve the accuracy of our detection method, we employ four novel features and capture their anomalies with the joint decisions. Extensive evaluations show that FLTracer achieves an average true positive rate of over $96.88\%$ at an average false positive rate of less than $2.67\%$, significantly outperforming SOTA detection methods. \footnote{Code is available at \url{https://github.com/Eyr3/FLTracer}.}

Jiachen Lei, Qinglong Wang, Peng Cheng et al.

Diffusion model has emerged as the \emph{de-facto} model for image generation, yet the heavy training overhead hinders its broader adoption in the research community. We observe that diffusion models are commonly trained to learn all fine-grained visual information from scratch. This paradigm may cause unnecessary training costs hence requiring in-depth investigation. In this work, we show that it suffices to train a strong diffusion model by first pre-training the model to learn some primer distribution that loosely characterizes the unknown real image distribution. Then the pre-trained model can be fine-tuned for various generation tasks efficiently. In the pre-training stage, we propose to mask a high proportion (e.g., up to 90\%) of input images to approximately represent the primer distribution and introduce a masked denoising score matching objective to train a model to denoise visible areas. In subsequent fine-tuning stage, we efficiently train diffusion model without masking. Utilizing the two-stage training framework, we achieves significant training acceleration and a new FID score record of 6.27 on CelebA-HQ $256 \times 256$ for ViT-based diffusion models. The generalizability of a pre-trained model further helps building models that perform better than ones trained from scratch on different downstream datasets. For instance, a diffusion model pre-trained on VGGFace2 attains a 46\% quality improvement when fine-tuned on a different dataset that contains only 3000 images. Our code is available at \url{https://github.com/jiachenlei/maskdm}.

Qingcao Li, Yipeng Lin, Weichen Lian et al.

Speech deepfake detection has achieved remarkable success in clean environments but faces significant challenges in complex, real-world scenarios where speech is often mixed with background music or noise. Current state-of-the-art methods rely on semantic features from self-supervised learning (SSL) models, which often fail when processing non-speech or mixed-source audio. In this paper, we first introduce MixFake, a large-scale benchmark dataset designed to simulate diverse acoustic environments with varying SNR levels and mixed authenticity components. To address the "semantic-centric" limitation, we propose a Multi-stream Prompt Tuning framework that injects signal-level priors into SSL backbones. By integrating base, frequency, and texture streams through deep prompt injection, our model effectively captures acoustic artifacts. Experimental results demonstrate that our method significantly outperforms existing baselines, achieving a 0.95% EER in foreground detection and a substantial 7.72% absolute improvement in complex background detection tasks. Our dataset and code are available at https://github.com/saltfish233/MixFake.

Hanbin Hong, Xinyu Zhang, Binghui Wang et al.

Black-box adversarial attacks have demonstrated strong potential to compromise machine learning models by iteratively querying the target model or leveraging transferability from a local surrogate model. Recently, such attacks can be effectively mitigated by state-of-the-art (SOTA) defenses, e.g., detection via the pattern of sequential queries, or injecting noise into the model. To our best knowledge, we take the first step to study a new paradigm of black-box attacks with provable guarantees -- certifiable black-box attacks that can guarantee the attack success probability (ASP) of adversarial examples before querying over the target model. This new black-box attack unveils significant vulnerabilities of machine learning models, compared to traditional empirical black-box attacks, e.g., breaking strong SOTA defenses with provable confidence, constructing a space of (infinite) adversarial examples with high ASP, and the ASP of the generated adversarial examples is theoretically guaranteed without verification/queries over the target model. Specifically, we establish a novel theoretical foundation for ensuring the ASP of the black-box attack with randomized adversarial examples (AEs). Then, we propose several novel techniques to craft the randomized AEs while reducing the perturbation size for better imperceptibility. Finally, we have comprehensively evaluated the certifiable black-box attacks on the CIFAR10/100, ImageNet, and LibriSpeech datasets, while benchmarking with 16 SOTA black-box attacks, against various SOTA defenses in the domains of computer vision and speech recognition. Both theoretical and experimental results have validated the significance of the proposed attack. The code and all the benchmarks are available at \url{https://github.com/datasec-lab/CertifiedAttack}.

Yao Wei, Yanchao Sun, Ruijie Zheng et al.

We introduce DualMind, a generalist agent designed to tackle various decision-making tasks that addresses challenges posed by current methods, such as overfitting behaviors and dependence on task-specific fine-tuning. DualMind uses a novel "Dual-phase" training strategy that emulates how humans learn to act in the world. The model first learns fundamental common knowledge through a self-supervised objective tailored for control tasks and then learns how to make decisions based on different contexts through imitating behaviors conditioned on given prompts. DualMind can handle tasks across domains, scenes, and embodiments using just a single set of model weights and can execute zero-shot prompting without requiring task-specific fine-tuning. We evaluate DualMind on MetaWorld and Habitat through extensive experiments and demonstrate its superior generalizability compared to previous techniques, outperforming other generalist agents by over 50$\%$ and 70$\%$ on Habitat and MetaWorld, respectively. On the 45 tasks in MetaWorld, DualMind achieves over 30 tasks at a 90$\%$ success rate.

Zhongjie Ba, Jieming Zhong, Jiachen Lei et al.

Advanced text-to-image models such as DALL$\cdot$E 2 and Midjourney possess the capacity to generate highly realistic images, raising significant concerns regarding the potential proliferation of unsafe content. This includes adult, violent, or deceptive imagery of political figures. Despite claims of rigorous safety mechanisms implemented in these models to restrict the generation of not-safe-for-work (NSFW) content, we successfully devise and exhibit the first prompt attacks on Midjourney, resulting in the production of abundant photorealistic NSFW images. We reveal the fundamental principles of such prompt attacks and suggest strategically substituting high-risk sections within a suspect prompt to evade closed-source safety measures. Our novel framework, SurrogatePrompt, systematically generates attack prompts, utilizing large language models, image-to-text, and image-to-image modules to automate attack prompt creation at scale. Evaluation results disclose an 88% success rate in bypassing Midjourney's proprietary safety filter with our attack prompts, leading to the generation of counterfeit images depicting political figures in violent scenarios. Both subjective and objective assessments validate that the images generated from our attack prompts present considerable safety hazards.

Xinyu Zhang, Hanbin Hong, Yuan Hong et al.

The language models, especially the basic text classification models, have been shown to be susceptible to textual adversarial attacks such as synonym substitution and word insertion attacks. To defend against such attacks, a growing body of research has been devoted to improving the model robustness. However, providing provable robustness guarantees instead of empirical robustness is still widely unexplored. In this paper, we propose Text-CRS, a generalized certified robustness framework for natural language processing (NLP) based on randomized smoothing. To our best knowledge, existing certified schemes for NLP can only certify the robustness against $\ell_0$ perturbations in synonym substitution attacks. Representing each word-level adversarial operation (i.e., synonym substitution, word reordering, insertion, and deletion) as a combination of permutation and embedding transformation, we propose novel smoothing theorems to derive robustness bounds in both permutation and embedding space against such adversarial operations. To further improve certified accuracy and radius, we consider the numerical relationships between discrete words and select proper noise distributions for the randomized smoothing. Finally, we conduct substantial experiments on multiple language models and datasets. Text-CRS can address all four different word-level adversarial operations and achieve a significant accuracy improvement. We also provide the first benchmark on certified accuracy and radius of four word-level operations, besides outperforming the state-of-the-art certification against synonym substitution attacks.

Jiachen Lei, Shuang Ma, Zhongjie Ba et al.

In this report, we present our approach and empirical results of applying masked autoencoders in two egocentric video understanding tasks, namely, Object State Change Classification and PNR Temporal Localization, of Ego4D Challenge 2022. As team TheSSVL, we ranked 2nd place in both tasks. Our code will be made available.

Huiyu Xu, Wenhui Zhang, Zhibo Wang et al.

Recently, advanced Large Language Models (LLMs) such as GPT-4 have been integrated into many real-world applications like Code Copilot. These applications have significantly expanded the attack surface of LLMs, exposing them to a variety of threats. Among them, jailbreak attacks that induce toxic responses through jailbreak prompts have raised critical safety concerns. To identify these threats, a growing number of red teaming approaches simulate potential adversarial scenarios by crafting jailbreak prompts to test the target LLM. However, existing red teaming methods do not consider the unique vulnerabilities of LLM in different scenarios, making it difficult to adjust the jailbreak prompts to find context-specific vulnerabilities. Meanwhile, these methods are limited to refining jailbreak templates using a few mutation operations, lacking the automation and scalability to adapt to different scenarios. To enable context-aware and efficient red teaming, we abstract and model existing attacks into a coherent concept called "jailbreak strategy" and propose a multi-agent LLM system named RedAgent that leverages these strategies to generate context-aware jailbreak prompts. By self-reflecting on contextual feedback in an additional memory buffer, RedAgent continuously learns how to leverage these strategies to achieve effective jailbreaks in specific contexts. Extensive experiments demonstrate that our system can jailbreak most black-box LLMs in just five queries, improving the efficiency of existing red teaming methods by two times. Additionally, RedAgent can jailbreak customized LLM applications more efficiently. By generating context-aware jailbreak prompts towards applications on GPTs, we discover 60 severe vulnerabilities of these real-world applications with only two queries per vulnerability. We have reported all found issues and communicated with OpenAI and Meta for bug fixes.

Peng Cheng, Yuwei Wang, Peng Huang et al.

Extensive research has revealed that adversarial examples (AE) pose a significant threat to voice-controllable smart devices. Recent studies have proposed black-box adversarial attacks that require only the final transcription from an automatic speech recognition (ASR) system. However, these attacks typically involve many queries to the ASR, resulting in substantial costs. Moreover, AE-based adversarial audio samples are susceptible to ASR updates. In this paper, we identify the root cause of these limitations, namely the inability to construct AE attack samples directly around the decision boundary of deep learning (DL) models. Building on this observation, we propose ALIF, the first black-box adversarial linguistic feature-based attack pipeline. We leverage the reciprocal process of text-to-speech (TTS) and ASR models to generate perturbations in the linguistic embedding space where the decision boundary resides. Based on the ALIF pipeline, we present the ALIF-OTL and ALIF-OTA schemes for launching attacks in both the digital domain and the physical playback environment on four commercial ASRs and voice assistants. Extensive evaluations demonstrate that ALIF-OTL and -OTA significantly improve query efficiency by 97.7% and 73.3%, respectively, while achieving competitive performance compared to existing methods. Notably, ALIF-OTL can generate an attack sample with only one query. Furthermore, our test-of-time experiment validates the robustness of our approach against ASR updates.

Xinyu Zhang, Ziping Dong, Qingyu Liu et al.

The rapid advancement of generative AI has underscored the critical need for identifying image ownership and protecting copyrights. This makes post-processing image watermarking an essential tool -- it involves embedding a specific watermark message into an image, with successful verification if a similar message can be decoded from the watermarked image. However, this method is susceptible to both adversarial attacks that manipulate the watermarked image to yield an unverified message upon decoding, and the proposed identity leakage-related attacks (e.g., forging watermarked images). The threat of identity leakage is particularly exacerbated in both empirical and certified robust watermarking methods. To defend against the aforementioned attacks, we propose W-IR, the first image watermarking framework that simultaneously incorporates identity protection and robustness. To enhance model robustness, we introduce a novel randomized smoothing technique as part of a robust watermarking, that offers certified robustness against perturbations across two distinct transformation spaces: pixel-level and coordinate-level. Moreover, to further mitigate identity leakage, we propose a new strategy based on residual information loss, aimed at minimizing the mutual information between the residual and watermarked images. Our work strikes a superior balance between robustness and identity leakage mitigation. Extensive experiments demonstrate that our W-IR framework achieves high certified accuracy for authenticity while effectively reducing identity leakage. \footnote{The code is available at https://github.com/holdrain/W-I-R.}

Binhao Ma, Tianhang Zheng, Hongsheng Hu et al.

Machine learning models trained on vast amounts of real or synthetic data often achieve outstanding predictive performance across various domains. However, this utility comes with increasing concerns about privacy, as the training data may include sensitive information. To address these concerns, machine unlearning has been proposed to erase specific data samples from models. While some unlearning techniques efficiently remove data at low costs, recent research highlights vulnerabilities where malicious users could request unlearning on manipulated data to compromise the model. Despite these attacks' effectiveness, perturbed data differs from original training data, failing hash verification. Existing attacks on machine unlearning also suffer from practical limitations and require substantial additional knowledge and resources. To fill the gaps in current unlearning attacks, we introduce the Unlearning Usability Attack. This model-agnostic, unlearning-agnostic, and budget-friendly attack distills data distribution information into a small set of benign data. These data are identified as benign by automatic poisoning detection tools due to their positive impact on model training. While benign for machine learning, unlearning these data significantly degrades model information. Our evaluation demonstrates that unlearning this benign data, comprising no more than 1% of the total training data, can reduce model accuracy by up to 50%. Furthermore, our findings show that well-prepared benign data poses challenges for recent unlearning techniques, as erasing these synthetic instances demands higher resources than regular data. These insights underscore the need for future research to reconsider "data poisoning" in the context of machine unlearning.

Meng Chen, Li Lu, Jiadi Yu et al.

Faced with the threat of identity leakage during voice data publishing, users are engaged in a privacy-utility dilemma when enjoying convenient voice services. Existing studies employ direct modification or text-based re-synthesis to de-identify users' voices, but resulting in inconsistent audibility in the presence of human participants. In this paper, we propose a voice de-identification system, which uses adversarial examples to balance the privacy and utility of voice services. Instead of typical additive examples inducing perceivable distortions, we design a novel convolutional adversarial example that modulates perturbations into real-world room impulse responses. Benefit from this, our system could preserve user identity from exposure by Automatic Speaker Identification (ASI) while remaining the voice perceptual quality for non-intrusive de-identification. Moreover, our system learns a compact speaker distribution through a conditional variational auto-encoder to sample diverse target embeddings on demand. Combining diverse target generation and input-specific perturbation construction, our system enables any-to-any identify transformation for adaptive de-identification. Experimental results show that our system could achieve 98% and 79% successful de-identification on mainstream ASIs and commercial systems with an objective Mel cepstral distortion of 4.31dB and a subjective mean opinion score of 4.48.

Zhongjie Ba, Qingyu Liu, Zhenguang Liu et al.

Deepfake technology has given rise to a spectrum of novel and compelling applications. Unfortunately, the widespread proliferation of high-fidelity fake videos has led to pervasive confusion and deception, shattering our faith that seeing is believing. One aspect that has been overlooked so far is that current deepfake detection approaches may easily fall into the trap of overfitting, focusing only on forgery clues within one or a few local regions. Moreover, existing works heavily rely on neural networks to extract forgery features, lacking theoretical constraints guaranteeing that sufficient forgery clues are extracted and superfluous features are eliminated. These deficiencies culminate in unsatisfactory accuracy and limited generalizability in real-life scenarios. In this paper, we try to tackle these challenges through three designs: (1) We present a novel framework to capture broader forgery clues by extracting multiple non-overlapping local representations and fusing them into a global semantic-rich feature. (2) Based on the information bottleneck theory, we derive Local Information Loss to guarantee the orthogonality of local representations while preserving comprehensive task-relevant information. (3) Further, to fuse the local representations and remove task-irrelevant information, we arrive at a Global Information Loss through the theoretical analysis of mutual information. Empirically, our method achieves state-of-the-art performance on five benchmark datasets.Our code is available at \url{https://github.com/QingyuLiu/Exposing-the-Deception}, hoping to inspire researchers.

Jianpeng Wang, Haoyu Wang, Baoying Chen et al.

The rapid democratization of prompt-based AI image editing has recently exacerbated the risks associated with malicious content fabrication and misinformation. However, forgery localization methods targeting these emerging editing techniques remain significantly under-explored. To bridge this gap, we first introduce a fully automated mask annotating framework that leverages keypoint alignment and semantic space similarity to generate precise ground-truth masks for edited regions. Based on this framework, we construct PromptForge-350k, a large-scale forgery localization dataset covering four state-of-the-art prompt-based AI image editing models, thereby mitigating the data scarcity in this domain. Furthermore, we propose ICL-Net, an effective forgery localization network featuring a triple-stream backbone and intra-image contrastive learning. This design enables the model to capture highly robust and generalizable forensic features. Extensive experiments demonstrate that our method achieves an IoU of 62.5% on PromptForge-350k, outperforming SOTA methods by 5.1%. Additionally, it exhibits strong robustness against common degradations with an IoU drop of less than 1%, and shows promising generalization capabilities on unseen editing models, achieving an average IoU of 41.5%.

Chao Shuai, Zhenguang Liu, Shaojing Fan et al.

AI-generated image detection has become increasingly important with the rapid advancement of generative AI. However, detectors built on Vision Foundation Models (VFMs, \emph{e.g.}, CLIP) often struggle to generalize to images created using unseen generation pipelines. We identify, for the first time, a key failure mechanism, termed \emph{semantic fallback}, where VFM-based detectors rely on dominant pre-trained semantic priors (such as identity) rather than forgery-specific traces under distribution shifts. To address this issue, we propose \textbf{Geometric Semantic Decoupling (GSD)}, a parameter-free module that explicitly removes semantic components from learned representations by leveraging a frozen VFM as a semantic guide with a trainable VFM as an artifact detector. GSD estimates semantic directions from batch-wise statistics and projects them out via a geometric constraint, forcing the artifact detector to rely on semantic-invariant forensic evidence. Extensive experiments demonstrate that our method consistently outperforms state-of-the-art approaches, achieving 94.4\% video-level AUC (+\textbf{1.2\%}) in cross-dataset evaluation, improving robustness to unseen manipulations (+\textbf{3.0\%} on DF40), and generalizing beyond faces to the detection of synthetic images of general scenes, including UniversalFakeDetect (+\textbf{0.9\%}) and GenImage (+\textbf{1.7\%}).

Dongrui Liu, Qihan Ren, Chen Qian et al.

The rise of AI agents introduces complex safety and security challenges arising from autonomous tool use and environmental interactions. Current guardrail models lack agentic risk awareness and transparency in risk diagnosis. To introduce an agentic guardrail that covers complex and numerous risky behaviors, we first propose a unified three-dimensional taxonomy that orthogonally categorizes agentic risks by their source (where), failure mode (how), and consequence (what). Guided by this structured and hierarchical taxonomy, we introduce a new fine-grained agentic safety benchmark (ATBench) and a Diagnostic Guardrail framework for agent safety and security (AgentDoG). AgentDoG provides fine-grained and contextual monitoring across agent trajectories. More Crucially, AgentDoG can diagnose the root causes of unsafe actions and seemingly safe but unreasonable actions, offering provenance and transparency beyond binary labels to facilitate effective agent alignment. AgentDoG variants are available in three sizes (4B, 7B, and 8B parameters) across Qwen and Llama model families. Extensive experimental results demonstrate that AgentDoG achieves state-of-the-art performance in agentic safety moderation in diverse and complex interactive scenarios. All models and datasets are openly released.

Qing Wen, Haohao Li, Zhongjie Ba et al.

Advances in AIGC technologies have enabled the synthesis of highly realistic audio deepfakes capable of deceiving human auditory perception. Although numerous audio deepfake detection (ADD) methods have been developed, most rely on local temporal/spectral features or pairwise relations, overlooking high-order interactions (HOIs). HOIs capture discriminative patterns that emerge from multiple feature components beyond their individual contributions. We propose HyperPotter, a hypergraph-based framework that explicitly models these synergistic HOIs through clustering-based hyperedges with class-aware prototype initialization. Extensive experiments demonstrate that HyperPotter surpasses its baseline by an average relative gain of 22.15% across 11 datasets and outperforms state-of-the-art methods by 13.96% on 4 challenging cross-domain datasets, demonstrating superior generalization to diverse attacks and speakers.

Yu He, Boheng Li, Liu Liu et al.

Membership Inference Attacks (MIAs) aim to predict whether a data sample belongs to the model's training set or not. Although prior research has extensively explored MIAs in Large Language Models (LLMs), they typically require accessing to complete output logits (\ie, \textit{logits-based attacks}), which are usually not available in practice. In this paper, we study the vulnerability of pre-trained LLMs to MIAs in the \textit{label-only setting}, where the adversary can only access generated tokens (text). We first reveal that existing label-only MIAs have minor effects in attacking pre-trained LLMs, although they are highly effective in inferring fine-tuning datasets used for personalized LLMs. We find that their failure stems from two main reasons, including better generalization and overly coarse perturbation. Specifically, due to the extensive pre-training corpora and exposing each sample only a few times, LLMs exhibit minimal robustness differences between members and non-members. This makes token-level perturbations too coarse to capture such differences. To alleviate these problems, we propose \textbf{PETAL}: a label-only membership inference attack based on \textbf{PE}r-\textbf{T}oken sem\textbf{A}ntic simi\textbf{L}arity. Specifically, PETAL leverages token-level semantic similarity to approximate output probabilities and subsequently calculate the perplexity. It finally exposes membership based on the common assumption that members are `better' memorized and have smaller perplexity. We conduct extensive experiments on the WikiMIA benchmark and the more challenging MIMIR benchmark. Empirically, our PETAL performs better than the extensions of existing label-only attacks against personalized LLMs and even on par with other advanced logit-based attacks across all metrics on five prevalent open-source LLMs.

Dadi Guo, Qingyu Liu, Dongrui Liu et al.

Large Language Model (LLM)-based agents are increasingly used as autonomous subordinates that carry out tasks for users. This raises the question of whether they may also engage in deception, similar to how individuals in human organizations lie to superiors to create a good image or avoid punishment. We observe and define agentic upward deception, a phenomenon in which an agent facing environmental constraints conceals its failure and performs actions that were not requested without reporting. To assess its prevalence, we construct a benchmark of 200 tasks covering five task types and eight realistic scenarios in a constrained environment, such as broken tools or mismatched information sources. Evaluations of 11 popular LLMs reveal that these agents typically exhibit action-based deceptive behaviors, such as guessing results, performing unsupported simulations, substituting unavailable information sources, and fabricating local files. We further test prompt-based mitigation and find only limited reductions, suggesting that it is difficult to eliminate and highlighting the need for stronger mitigation strategies to ensure the safety of LLM-based agents.

Zhongjie Ba, Liang Yi, Peng Cheng et al.

Toxic speech detection has become a crucial challenge in maintaining safe online communication environments. However, existing approaches to toxic speech detection often neglect the contribution of paralinguistic cues, such as emotion, intonation, and speech rate, which are key to detecting speech toxicity. Moreover, current toxic speech datasets are predominantly text-based, limiting the development of models that can capture paralinguistic cues.To address these challenges, we present ToxiAlert-Bench, a large-scale audio dataset comprising over 30,000 audio clips annotated with seven major toxic categories and twenty fine-grained toxic labels. Uniquely, our dataset annotates toxicity sources -- distinguishing between textual content and paralinguistic origins -- for comprehensive toxic speech analysis.Furthermore, we propose a dual-head neural network with a multi-stage training strategy tailored for toxic speech detection. This architecture features two task-specific classification headers: one for identifying the source of sensitivity (textual or paralinguistic), and the other for categorizing the specific toxic type. The training process involves independent head training followed by joint fine-tuning to reduce task interference. To mitigate data class imbalance, we incorporate class-balanced sampling and weighted loss functions.Our experimental results show that leveraging paralinguistic features significantly improves detection performance. Our method consistently outperforms existing baselines across multiple evaluation metrics, with a 21.1% relative improvement in Macro-F1 score and a 13.0% relative gain in accuracy over the strongest baseline, highlighting its enhanced effectiveness and practical applicability.

Ziping Dong, Chao Shuai, Zhongjie Ba et al.

Invisible Image Watermarking is crucial for ensuring content provenance and accountability in generative AI. While Gen-AI providers are increasingly integrating invisible watermarking systems, the robustness of these schemes against forgery attacks remains poorly characterized. This is critical, as forging traceable watermarks onto illicit content leads to false attribution, potentially harming the reputation and legal standing of Gen-AI service providers who are not responsible for the content. In this work, we propose WMCopier, an effective watermark forgery attack that operates without requiring any prior knowledge of or access to the target watermarking algorithm. Our approach first models the target watermark distribution using an unconditional diffusion model, and then seamlessly embeds the target watermark into a non-watermarked image via a shallow inversion process. We also incorporate an iterative optimization procedure that refines the reconstructed image to further trade off the fidelity and forgery efficiency. Experimental results demonstrate that WMCopier effectively deceives both open-source and closed-source watermark systems (e.g., Amazon's system), achieving a significantly higher success rate than existing methods. Additionally, we evaluate the robustness of forged samples and discuss the potential defenses against our attack.

Xiaojing Chen, Dan Li, Lijun Peng et al.

The prosperous development of Artificial Intelligence-Generated Content (AIGC) has brought people's anxiety about the spread of false information on social media. Designing detectors for filtering is an effective defense method, but most detectors will be compromised by adversarial samples. Currently, most studies exposing AIGC security issues assume information on model structure and data distribution. In real applications, attackers query and interfere with models that provide services in the form of application programming interfaces (APIs), which constitutes the black-box decision-based attack paradigm. However, to the best of our knowledge, decision-based attacks on AIGC detectors remain unexplored. In this study, we propose \textbf{FBA$^2$D}: a frequency-based black-box attack method for AIGC detection to fill the research gap. Motivated by frequency-domain discrepancies between generated and real images, we develop a decision-based attack that leverages the Discrete Cosine Transform (DCT) for fine-grained spectral partitioning and selects frequency bands as query subspaces, improving both query efficiency and image quality. Moreover, attacks on AIGC detectors should mitigate initialization failures, preserve image quality, and operate under strict query budgets. To address these issues, we adopt an ``adversarial example soup'' method, averaging candidates from successive surrogate iterations and using the result as the initialization to accelerate the query-based attack. The empirical study on the Synthetic LSUN dataset and GenImage dataset demonstrate the effectiveness of our prosed method. This study shows the urgency of addressing practical AIGC security problems.

Gaojian Wang, Feng Lin, Tong Wu et al.

This work asks: with abundant, unlabeled real faces, how to learn a robust and transferable facial representation that boosts various face security tasks with respect to generalization performance? We make the first attempt and propose a self-supervised pretraining framework to learn fundamental representations of real face images, FSFM, that leverages the synergy between masked image modeling (MIM) and instance discrimination (ID). We explore various facial masking strategies for MIM and present a simple yet powerful CRFR-P masking, which explicitly forces the model to capture meaningful intra-region consistency and challenging inter-region coherency. Furthermore, we devise the ID network that naturally couples with MIM to establish underlying local-to-global correspondence via tailored self-distillation. These three learning objectives, namely 3C, empower encoding both local features and global semantics of real faces. After pretraining, a vanilla ViT serves as a universal vision foundation model for downstream face security tasks: cross-dataset deepfake detection, cross-domain face anti-spoofing, and unseen diffusion facial forgery detection. Extensive experiments on 10 public datasets demonstrate that our model transfers better than supervised pretraining, visual and facial self-supervised learning arts, and even outperforms task-specialized SOTA methods.

Qingchao Jiang, Zhishuo Xu, Zhiying Zhu et al.

Advances in image generation enable hyper-realistic synthetic faces but also pose risks, thus making synthetic face detection crucial. Previous research focuses on the general differences between generated images and real images, often overlooking the discrepancies among various generative techniques. In this paper, we explore the intrinsic relationship between synthetic images and their corresponding generation technologies. We find that specific images exhibit significant reconstruction discrepancies across different generative methods and that matching generation techniques provide more accurate reconstructions. Based on this insight, we propose a Multi-Reconstruction-based detector. By reversing and reconstructing images using multiple generative models, we analyze the reconstruction differences among real, GAN-generated, and DM-generated images to facilitate effective differentiation. Additionally, we introduce the Asian Synthetic Face Dataset (ASFD), containing synthetic Asian faces generated with various GANs and DMs. This dataset complements existing synthetic face datasets. Experimental results demonstrate that our detector achieves exceptional performance, with strong generalization and robustness.

Zhenguang Liu, Chao Shuai, Shaojing Fan et al.

Diffusion models have achieved remarkable success in novel view synthesis, but their reliance on large, diverse, and often untraceable Web datasets has raised pressing concerns about image copyright protection. Current methods fall short in reliably identifying unauthorized image use, as they struggle to generalize across varied generation tasks and fail when the training dataset includes images from multiple sources with few identifiable (watermarked or poisoned) samples. In this paper, we present novel evidence that diffusion-generated images faithfully preserve the statistical properties of their training data, particularly reflected in their spectral features. Leveraging this insight, we introduce \emph{CoprGuard}, a robust frequency domain watermarking framework to safeguard against unauthorized image usage in diffusion model training and fine-tuning. CoprGuard demonstrates remarkable effectiveness against a wide range of models, from naive diffusion models to sophisticated text-to-image models, and is robust even when watermarked images comprise a mere 1\% of the training dataset. This robust and versatile approach empowers content owners to protect their intellectual property in the era of AI-driven image generation.

Zhongjie Ba, Yitao Zhang, Peng Cheng et al.

Watermarking plays a key role in the provenance and detection of AI-generated content. While existing methods prioritize robustness against real-world distortions (e.g., JPEG compression and noise addition), we reveal a fundamental tradeoff: such robust watermarks inherently improve the redundancy of detectable patterns encoded into images, creating exploitable information leakage. To leverage this, we propose an attack framework that extracts leakage of watermark patterns through multi-channel feature learning using a pre-trained vision model. Unlike prior works requiring massive data or detector access, our method achieves both forgery and detection evasion with a single watermarked image. Extensive experiments demonstrate that our method achieves a 60\% success rate gain in detection evasion and 51\% improvement in forgery accuracy compared to state-of-the-art methods while maintaining visual fidelity. Our work exposes the robustness-stealthiness paradox: current "robust" watermarks sacrifice security for distortion resistance, providing insights for future watermark design.

Sifan Wu, Zhenguang Liu, Beibei Zhang et al.

Human motion copy is an intriguing yet challenging task in artificial intelligence and computer vision, which strives to generate a fake video of a target person performing the motion of a source person. The problem is inherently challenging due to the subtle human-body texture details to be generated and the temporal consistency to be considered. Existing approaches typically adopt a conventional GAN with an L1 or L2 loss to produce the target fake video, which intrinsically necessitates a large number of training samples that are challenging to acquire. Meanwhile, current methods still have difficulties in attaining realistic image details and temporal consistency, which unfortunately can be easily perceived by human observers. Motivated by this, we try to tackle the issues from three aspects: (1) We constrain pose-to-appearance generation with a perceptual loss and a theoretically motivated Gromov-Wasserstein loss to bridge the gap between pose and appearance. (2) We present an episodic memory module in the pose-to-appearance generation to propel continuous learning that helps the model learn from its past poor generations. We also utilize geometrical cues of the face to optimize facial details and refine each key body part with a dedicated local GAN. (3) We advocate generating the foreground in a sequence-to-sequence manner rather than a single-frame manner, explicitly enforcing temporal inconsistency. Empirical results on five datasets, iPER, ComplexMotion, SoloDance, Fish, and Mouse datasets, demonstrate that our method is capable of generating realistic target videos while precisely copying motion from a source video. Our method significantly outperforms state-of-the-art approaches and gains 7.2% and 12.4% improvements in PSNR and FID respectively.