Tianwei Zhang

Shengyu Zhang, Linfeng Dong, Xiaoya Li et al.

This paper surveys research works in the quickly advancing field of instruction tuning (IT), which can also be referred to as supervised fine-tuning (SFT)\footnote{In this paper, unless specified otherwise, supervised fine-tuning (SFT) and instruction tuning (IT) are used interchangeably.}, a crucial technique to enhance the capabilities and controllability of large language models (LLMs). Instruction tuning refers to the process of further training LLMs on a dataset consisting of \textsc{(instruction, output)} pairs in a supervised fashion, which bridges the gap between the next-word prediction objective of LLMs and the users' objective of having LLMs adhere to human instructions. In this work, we make a systematic review of the literature, including the general methodology of SFT, the construction of SFT datasets, the training of SFT models, and applications to different modalities, domains and application, along with analysis on aspects that influence the outcome of SFT (e.g., generation of instruction outputs, size of the instruction dataset, etc). We also review the potential pitfalls of SFT along with criticism against it, along with efforts pointing out current deficiencies of existing strategies and suggest some avenues for fruitful research. Project Page: github.com/xiaoya-li/Instruction-Tuning-Survey

Boheng Li, Yanhao Wei, Yankai Fu et al.

Text-to-image diffusion models are pushing the boundaries of what generative AI can achieve in our lives. Beyond their ability to generate general images, new personalization techniques have been proposed to customize the pre-trained base models for crafting images with specific themes or styles. Such a lightweight solution, enabling AI practitioners and developers to easily build their own personalized models, also poses a new concern regarding whether the personalized models are trained from unauthorized data. A promising solution is to proactively enable data traceability in generative models, where data owners embed external coatings (e.g., image watermarks or backdoor triggers) onto the datasets before releasing. Later the models trained over such datasets will also learn the coatings and unconsciously reproduce them in the generated mimicries, which can be extracted and used as the data usage evidence. However, we identify the existing coatings cannot be effectively learned in personalization tasks, making the corresponding verification less reliable. In this paper, we introduce SIREN, a novel methodology to proactively trace unauthorized data usage in black-box personalized text-to-image diffusion models. Our approach optimizes the coating in a delicate way to be recognized by the model as a feature relevant to the personalization task, thus significantly improving its learnability. We also utilize a human perceptual-aware constraint, a hypersphere classification technique, and a hypothesis-testing-guided verification method to enhance the stealthiness and detection accuracy of the coating. The effectiveness of SIREN is verified through extensive experiments on a diverse set of benchmark datasets, models, and learning algorithms. SIREN is also effective in various real-world scenarios and evaluated against potential countermeasures. Our code is publicly available.

Guanlin Li, Shuai Yang, Jie Zhang et al.

With the development of generative models, the quality of generated content keeps increasing. Recently, open-source models have made it surprisingly easy to manipulate and edit photos and videos, with just a few simple prompts. While these cutting-edge technologies have gained popularity, they have also given rise to concerns regarding the privacy and portrait rights of individuals. Malicious users can exploit these tools for deceptive or illegal purposes. Although some previous works focus on protecting photos against generative models, we find there are still gaps between protecting videos and images in the aspects of efficiency and effectiveness. Therefore, we introduce our protection method, PRIME, to significantly reduce the time cost and improve the protection performance. Moreover, to evaluate our proposed protection method, we consider both objective metrics and human subjective metrics. Our evaluation results indicate that PRIME only costs 8.3% GPU hours of the cost of the previous state-of-the-art method and achieves better protection results on both human evaluation and objective metrics. Code can be found in https://github.com/GuanlinLee/prime.

Cheng Wang, Gelei Deng, Xianglin Yang et al.

Large Audio-Language Models (LALMs) are enhanced with audio perception capabilities, enabling them to effectively process and understand multimodal inputs that combine audio and text. However, their performance in handling conflicting information between audio and text modalities remains largely unexamined. This paper introduces MCR-BENCH, the first comprehensive benchmark specifically designed to evaluate how LALMs prioritize information when presented with inconsistent audio-text pairs. Through extensive evaluation across diverse audio understanding tasks, we reveal a concerning phenomenon: when inconsistencies exist between modalities, LALMs display a significant bias toward textual input, frequently disregarding audio evidence. This tendency leads to substantial performance degradation in audio-centric tasks and raises important reliability concerns for real-world applications. We further investigate the influencing factors of text bias, and explore mitigation strategies through supervised finetuning, and analyze model confidence patterns that reveal persistent overconfidence even with contradictory inputs. These findings underscore the need for improved modality balance during training and more sophisticated fusion mechanisms to enhance the robustness when handling conflicting multi-modal inputs. The project is available at https://github.com/WangCheng0116/MCR-BENCH.

Wei Gao, Xinyu Zhou, Peng Sun et al.

Key-Value cache (\texttt{KV} \texttt{cache}) compression has emerged as a promising technique to optimize Large Language Model (LLM) serving. It primarily decreases the memory consumption of \texttt{KV} \texttt{cache} to reduce the computation cost. Despite the development of many compression algorithms, their applications in production environments are still not prevalent. In this paper, we revisit mainstream \texttt{KV} \texttt{cache} compression solutions from a practical perspective. Our contributions are three-fold. First, we comprehensively review existing algorithmic designs and benchmark studies for \texttt{KV} \texttt{cache} compression and identify missing pieces in their performance measurement, which could hinder their adoption in practice. Second, we empirically evaluate representative \texttt{KV} \texttt{cache} compression methods to uncover two key issues that affect the computational efficiency: (1) while compressing \texttt{KV} \texttt{cache} can reduce memory consumption, current implementations (e.g., FlashAttention, PagedAttention) do not optimize for production-level LLM serving, resulting in suboptimal throughput performance; (2) compressing \texttt{KV} \texttt{cache} may lead to longer outputs, resulting in increased end-to-end latency. We further investigate the accuracy performance of individual samples rather than the overall performance, revealing the intrinsic limitations in \texttt{KV} \texttt{cache} compression when handling specific LLM tasks. Third, we provide tools to shed light on future \texttt{KV} \texttt{cache} compression studies and facilitate their practical deployment in production. They are open-sourced in \href{https://github.com/LLMkvsys/rethink-kv-compression}{https://github.com/LLMkvsys/rethink-kv-compression}.

Yukun Chen, Boheng Li, Yu Yuan et al.

Knowledge distillation (KD) is a vital technique for deploying deep neural networks (DNNs) on resource-constrained devices by transferring knowledge from large teacher models to lightweight student models. While teacher models from third-party platforms may undergo security verification (\eg, backdoor detection), we uncover a novel and critical threat: distillation-conditional backdoor attacks (DCBAs). DCBA injects dormant and undetectable backdoors into teacher models, which become activated in student models via the KD process, even with clean distillation datasets. While the direct extension of existing methods is ineffective for DCBA, we implement this attack by formulating it as a bilevel optimization problem and proposing a simple yet effective method (\ie, SCAR). Specifically, the inner optimization simulates the KD process by optimizing a surrogate student model, while the outer optimization leverages outputs from this surrogate to optimize the teacher model for implanting the conditional backdoor. Our SCAR addresses this complex optimization utilizing an implicit differentiation algorithm with a pre-optimized trigger injection function. Extensive experiments across diverse datasets, model architectures, and KD techniques validate the effectiveness of our SCAR and its resistance against existing backdoor detection, highlighting a significant yet previously overlooked vulnerability in the KD process. Our code is available at https://github.com/WhitolfChen/SCAR.

Wenjie Li, Siying Gu, Yiming Li et al.

Backdoor detection is currently the mainstream defense against backdoor attacks in federated learning (FL), where malicious clients upload poisoned updates that compromise the global model and undermine the reliability of FL deployments. Existing backdoor detection techniques fall into two categories, including passive and proactive ones, depending on whether the server proactively modifies the global model. However, both have inherent limitations in practice: passive defenses are vulnerable to common non-i.i.d. data distributions and random participation of FL clients, whereas current proactive defenses suffer inevitable out-of-distribution (OOD) bias because they rely on backdoor co-existence effects. To address these issues, we introduce a new proactive defense, dubbed Coward, inspired by our discovery of multi-backdoor collision effects, in which consecutively planted, distinct backdoors significantly suppress earlier ones. In general, we detect attackers by evaluating whether the server-injected, conflicting global watermark is erased during local training rather than retained. Our method preserves the advantages of proactive defenses in handling data heterogeneity (\ie, non-i.i.d. data) while mitigating the adverse impact of OOD bias through a revised detection mechanism. Extensive experiments on benchmark datasets confirm the effectiveness of Coward and its resilience to potential adaptive attacks. The code for our method would be available at https://github.com/still2009/cowardFL.

Yanzhou Li, Tianlin Li, Kangjie Chen et al.

Mainstream backdoor attack methods typically demand substantial tuning data for poisoning, limiting their practicality and potentially degrading the overall performance when applied to Large Language Models (LLMs). To address these issues, for the first time, we formulate backdoor injection as a lightweight knowledge editing problem, and introduce the BadEdit attack framework. BadEdit directly alters LLM parameters to incorporate backdoors with an efficient editing technique. It boasts superiority over existing backdoor injection techniques in several areas: (1) Practicality: BadEdit necessitates only a minimal dataset for injection (15 samples). (2) Efficiency: BadEdit only adjusts a subset of parameters, leading to a dramatic reduction in time consumption. (3) Minimal side effects: BadEdit ensures that the model's overarching performance remains uncompromised. (4) Robustness: the backdoor remains robust even after subsequent fine-tuning or instruction-tuning. Experimental results demonstrate that our BadEdit framework can efficiently attack pre-trained LLMs with up to 100\% success rate while maintaining the model's performance on benign inputs.

Haodong Li, Gelei Deng, Yi Liu et al.

Pre-training, which utilizes extensive and varied datasets, is a critical factor in the success of Large Language Models (LLMs) across numerous applications. However, the detailed makeup of these datasets is often not disclosed, leading to concerns about data security and potential misuse. This is particularly relevant when copyrighted material, still under legal protection, is used inappropriately, either intentionally or unintentionally, infringing on the rights of the authors. In this paper, we introduce a detailed framework designed to detect and assess the presence of content from potentially copyrighted books within the training datasets of LLMs. This framework also provides a confidence estimation for the likelihood of each content sample's inclusion. To validate our approach, we conduct a series of simulated experiments, the results of which affirm the framework's effectiveness in identifying and addressing instances of content misuse in LLM training processes. Furthermore, we investigate the presence of recognizable quotes from famous literary works within these datasets. The outcomes of our study have significant implications for ensuring the ethical use of copyrighted materials in the development of LLMs, highlighting the need for more transparent and responsible data management practices in this field.

Shiqian Zhao, Jiayang Liu, Yiming Li et al. · mit

Modern text-to-image (T2I) generation systems (e.g., DALL$\cdot$E 3) exploit the memory mechanism, which captures key information in multi-turn interactions for faithful generation. Despite its practicality, the security analyses of this mechanism have fallen far behind. In this paper, we reveal that it can exacerbate the risk of jailbreak attacks. Previous attacks fuse the unsafe target prompt into one ultimate adversarial prompt, which can be easily detected or lead to the generation of non-unsafe images due to under- or over-detoxification. In contrast, we propose embedding the malice at the inception of the chat session in memory, addressing the above limitations. Specifically, we propose Inception, the first multi-turn jailbreak attack against real-world text-to-image generation systems that explicitly exploits their memory mechanisms. Inception is composed of two key modules: segmentation and recursion. We introduce Segmentation, a semantic-preserving method that generates multi-round prompts. By leveraging NLP analysis techniques, we design policies to decompose a prompt, together with its malicious intent, according to sentence structure, thereby evading safety filters. Recursion further addresses the challenge posed by unsafe sub-prompts that cannot be separated through simple segmentation. It firstly expands the sub-prompt, then invokes segmentation recursively. To facilitate multi-turn adversarial prompts crafting, we build VisionFlow, an emulation T2I system that integrates two-stage safety filters and industrial-grade memory mechanisms. The experiment results show that Inception successfully allures unsafe image generation, surpassing the SOTA by a 20.0\% margin in attack success rate. We also conduct experiments on the real-world commercial T2I generation platforms, further validating the threats of Inception in practice.

Runyi Hu, Jie Zhang, Yiming Li et al.

In today's digital landscape, the blending of AI-generated and authentic content has underscored the need for copyright protection and content authentication. Watermarking has become a vital tool to address these challenges, safeguarding both generated and real content. Effective watermarking methods must withstand various distortions and attacks. Current deep watermarking techniques often use an encoder-noise layer-decoder architecture and include distortions to enhance robustness. However, they struggle to balance robustness and fidelity and remain vulnerable to adaptive attacks, despite extensive training. To overcome these limitations, we propose SuperMark, a robust, training-free watermarking framework. Inspired by the parallels between watermark embedding/extraction in watermarking and the denoising/noising processes in diffusion models, SuperMark embeds the watermark into initial Gaussian noise using existing techniques. It then applies pre-trained Super-Resolution (SR) models to denoise the watermarked noise, producing the final watermarked image. For extraction, the process is reversed: the watermarked image is inverted back to the initial watermarked noise via DDIM Inversion, from which the embedded watermark is extracted. This flexible framework supports various noise injection methods and diffusion-based SR models, enabling enhanced customization. The robustness of the DDIM Inversion process against perturbations allows SuperMark to achieve strong resilience to distortions while maintaining high fidelity. Experiments demonstrate that SuperMark achieves fidelity comparable to existing methods while significantly improving robustness. Under standard distortions, it achieves an average watermark extraction accuracy of 99.46%, and 89.29% under adaptive attacks. Moreover, SuperMark shows strong transferability across datasets, SR models, embedding methods, and resolutions.

Haoran Ou, Kangjie Chen, Xingshuo Han et al.

Large Language Models (LLMs) excel at tasks such as dialogue, summarization, and question answering, yet they struggle to adapt to specialized domains and evolving facts. To overcome this, web search has been integrated into LLMs, allowing real-time access to online content. However, this connection magnifies safety risks, as adversarial prompts combined with untrusted sources can cause severe vulnerabilities. We investigate red teaming for LLMs with web search and present CREST-Search, a framework that systematically exposes risks in such systems. Unlike existing methods for standalone LLMs, CREST-Search addresses the complex workflow of search-enabled models by generating adversarial queries with in-context learning and refining them through iterative feedback. We further construct WebSearch-Harm, a search-specific dataset to fine-tune LLMs into efficient red-teaming agents. Experiments show that CREST-Search effectively bypasses safety filters and reveals vulnerabilities in modern web-augmented LLMs, underscoring the need for specialized defenses to ensure trustworthy deployment.

Boheng Li, Junjie Wang, Yiming Li et al.

Despite the integration of safety alignment and external filters, text-to-image (T2I) generative models are still susceptible to producing harmful content, such as sexual or violent imagery. This raises serious concerns about unintended exposure and potential misuse. Red teaming, which aims to proactively identify diverse prompts that can elicit unsafe outputs from the T2I system (including the core generative model as well as potential external safety filters and other processing components), is increasingly recognized as an essential method for assessing and improving safety before real-world deployment. Yet, existing automated red teaming approaches often treat prompt discovery as an isolated, prompt-level optimization task, which limits their scalability, diversity, and overall effectiveness. To bridge this gap, in this paper, we propose DREAM, a scalable red teaming framework to automatically uncover diverse problematic prompts from a given T2I system. Unlike most prior works that optimize prompts individually, DREAM directly models the probabilistic distribution of the target system's problematic prompts, which enables explicit optimization over both effectiveness and diversity, and allows efficient large-scale sampling after training. To achieve this without direct access to representative training samples, we draw inspiration from energy-based models and reformulate the objective into simple and tractable objectives. We further introduce GC-SPSA, an efficient optimization algorithm that provide stable gradient estimates through the long and potentially non-differentiable T2I pipeline. The effectiveness of DREAM is validated through extensive experiments, demonstrating that it surpasses 9 state-of-the-art baselines by a notable margin across a broad range of T2I models and safety filters in terms of prompt success rate and diversity.

Yi Liu, Gelei Deng, Zhengzi Xu et al.

Large Language Models (LLMs), like ChatGPT, have demonstrated vast potential but also introduce challenges related to content constraints and potential misuse. Our study investigates three key research questions: (1) the number of different prompt types that can jailbreak LLMs, (2) the effectiveness of jailbreak prompts in circumventing LLM constraints, and (3) the resilience of ChatGPT against these jailbreak prompts. Initially, we develop a classification model to analyze the distribution of existing prompts, identifying ten distinct patterns and three categories of jailbreak prompts. Subsequently, we assess the jailbreak capability of prompts with ChatGPT versions 3.5 and 4.0, utilizing a dataset of 3,120 jailbreak questions across eight prohibited scenarios. Finally, we evaluate the resistance of ChatGPT against jailbreak prompts, finding that the prompts can consistently evade the restrictions in 40 use-case scenarios. The study underscores the importance of prompt structures in jailbreaking LLMs and discusses the challenges of robust jailbreak prompt generation and prevention.

Shangwei Guo, Chunlong Xie, Jiwei Li et al.

Pre-trained language models (PTLMs) have achieved great success and remarkable performance over a wide range of natural language processing (NLP) tasks. However, there are also growing concerns regarding the potential security issues in the adoption of PTLMs. In this survey, we comprehensively systematize recently discovered threats to PTLM systems and applications. We perform our attack characterization from three interesting perspectives. (1) We show threats can occur at different stages of the PTLM pipeline raised by different malicious entities. (2) We identify two types of model transferability (landscape, portrait) that facilitate attacks. (3) Based on the attack goals, we summarize four categories of attacks (backdoor, evasion, data privacy and model privacy). We also discuss some open problems and research directions. We believe our survey and taxonomy will inspire future studies towards secure and privacy-preserving PTLMs.

Xiaoxuan Lou, Tianwei Zhang, Jun Jiang et al.

Side-channel attacks have become a severe threat to the confidentiality of computer applications and systems. One popular type of such attacks is the microarchitectural attack, where the adversary exploits the hardware features to break the protection enforced by the operating system and steal the secrets from the program. In this paper, we systematize microarchitectural side channels with a focus on attacks and defenses in cryptographic applications. We make three contributions. (1) We survey past research literature to categorize microarchitectural side-channel attacks. Since these are hardware attacks targeting software, we summarize the vulnerable implementations in software, as well as flawed designs in hardware. (2) We identify common strategies to mitigate microarchitectural attacks, from the application, OS and hardware levels. (3) We conduct a large-scale evaluation on popular cryptographic applications in the real world, and analyze the severity, practicality and impact of side-channel vulnerabilities. This survey is expected to inspire side-channel research community to discover new attacks, and more importantly, propose new defense solutions against them.

Tianwei Zhang, Jun Jiang, Yinqian Zhang

We systematize software side-channel attacks with a focus on vulnerabilities and countermeasures in the cryptographic implementations. Particularly, we survey past research literature to categorize vulnerable implementations, and identify common strategies to eliminate them. We then evaluate popular libraries and applications, quantitatively measuring and comparing the vulnerability severity, response time and coverage. Based on these characterizations and evaluations, we offer some insights for side-channel researchers, cryptographic software developers and users. We hope our study can inspire the side-channel research community to discover new vulnerabilities, and more importantly, to fortify applications against them.

Jakub Szefer, Tianwei Zhang, Ruby B. Lee

We present a new and practical framework for security verification of secure architectures. Specifically, we break the verification task into external verification and internal verification. External verification considers the external protocols, i.e. interactions between users, compute servers, network entities, etc. Meanwhile, internal verification considers the interactions between hardware and software components within each server. This verification framework is general-purpose and can be applied to a stand-alone server, or a large-scale distributed system. We evaluate our verification method on the CloudMonatt and HyperWall architectures as examples.

Tianwei Zhang, Yinqian Zhang, Ruby B. Lee

In cloud computing, network Denial of Service (DoS) attacks are well studied and defenses have been implemented, but severe DoS attacks on a victim's working memory by a single hostile VM are not well understood. Memory DoS attacks are Denial of Service (or Degradation of Service) attacks caused by contention for hardware memory resources on a cloud server. Despite the strong memory isolation techniques for virtual machines (VMs) enforced by the software virtualization layer in cloud servers, the underlying hardware memory layers are still shared by the VMs and can be exploited by a clever attacker in a hostile VM co-located on the same server as the victim VM, denying the victim the working memory he needs. We first show quantitatively the severity of contention on different memory resources. We then show that a malicious cloud customer can mount low-cost attacks to cause severe performance degradation for a Hadoop distributed application, and 38X delay in response time for an E-commerce website in the Amazon EC2 cloud. Then, we design an effective, new defense against these memory DoS attacks, using a statistical metric to detect their existence and execution throttling to mitigate the attack damage. We achieve this by a novel re-purposing of existing hardware performance counters and duty cycle modulation for security, rather than for improving performance or power consumption. We implement a full prototype on the OpenStack cloud system. Our evaluations show that this defense system can effectively defeat memory DoS attacks with negligible performance overhead.