A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deployment

The remarkable success of Large Language Models (LLMs) has illuminated a promising pathway toward achieving Artificial General Intelligence for both academic and industrial communities, owing to their unprecedented performance across various applications. As LLMs continue to gain prominence in both research and commercial domains, their security and safety implications have become a growing concern, not only for researchers and corporations but also for every nation. Currently, existing surveys on LLM safety primarily focus on specific stages of the LLM lifecycle, e.g., deployment phase or fine-tuning phase, lacking a comprehensive understanding of the entire "lifechain" of LLMs. To address this gap, this paper introduces, for the first time, the concept of "full-stack" safety to systematically consider safety issues throughout the entire process of LLM training, deployment, and eventual commercialization. Compared to the off-the-shelf LLM safety surveys, our work demonstrates several distinctive advantages: (I) Comprehensive Perspective. We define the complete LLM lifecycle as encompassing data preparation, pre-training, post-training, deployment and final commercialization. To our knowledge, this represents the first safety survey to encompass the entire lifecycle of LLMs. (II) Extensive Literature Support. Our research is grounded in an exhaustive review of over 800+ papers, ensuring comprehensive coverage and systematic organization of security issues within a more holistic understanding. (III) Unique Insights. Through systematic literature analysis, we have developed reliable roadmaps and perspectives for each chapter. Our work identifies promising research directions, including safety in data generation, alignment techniques, model editing, and LLM-based agent systems. These insights provide valuable guidance for researchers pursuing future work in this field.

Privacy Protection Against Personalized Text-to-Image Synthesis via Cross-image Consistency Constraints

The rapid advancement of diffusion models and personalization techniques has made it possible to recreate individual portraits from just a few publicly available images. While such capabilities empower various creative applications, they also introduce serious privacy concerns, as adversaries can exploit them to generate highly realistic impersonations. To counter these threats, anti-personalization methods have been proposed, which add adversarial perturbations to published images to disrupt the training of personalization models. However, existing approaches largely overlook the intrinsic multi-image nature of personalization and instead adopt a naive strategy of applying perturbations independently, as commonly done in single-image settings. This neglects the opportunity to leverage inter-image relationships for stronger privacy protection. Therefore, we advocate for a group-level perspective on privacy protection against personalization. Specifically, we introduce Cross-image Anti-Personalization (CAP), a novel framework that enhances resistance to personalization by enforcing style consistency across perturbed images. Furthermore, we develop a dynamic ratio adjustment strategy that adaptively balances the impact of the consistency loss throughout the attack iterations. Extensive experiments on the classical CelebHQ and VGGFace2 benchmarks show that CAP substantially improves existing methods.

PATFinger: Prompt-Adapted Transferable Fingerprinting against Unauthorized Multimodal Dataset Usage

The multimodal datasets can be leveraged to pre-train large-scale vision-language models by providing cross-modal semantics. Current endeavors for determining the usage of datasets mainly focus on single-modal dataset ownership verification through intrusive methods and non-intrusive techniques, while cross-modal approaches remain under-explored. Intrusive methods can adapt to multimodal datasets but degrade model accuracy, while non-intrusive methods rely on label-driven decision boundaries that fail to guarantee stable behaviors for verification. To address these issues, we propose a novel prompt-adapted transferable fingerprinting scheme from a training-free perspective, called PATFinger, which incorporates the global optimal perturbation (GOP) and the adaptive prompts to capture dataset-specific distribution characteristics. Our scheme utilizes inherent dataset attributes as fingerprints instead of compelling the model to learn triggers. The GOP is derived from the sample distribution to maximize embedding drifts between different modalities. Subsequently, our PATFinger re-aligns the adaptive prompt with GOP samples to capture the cross-modal interactions on the carefully crafted surrogate model. This allows the dataset owner to check the usage of datasets by observing specific prediction behaviors linked to the PATFinger during retrieval queries. Extensive experiments demonstrate the effectiveness of our scheme against unauthorized multimodal dataset usage on various cross-modal retrieval architectures by 30% over state-of-the-art baselines.

Evolution-based Region Adversarial Prompt Learning for Robustness Enhancement in Vision-Language Models

Large pre-trained vision-language models (VLMs), such as CLIP, demonstrate impressive generalization but remain highly vulnerable to adversarial examples (AEs). Previous work has explored robust text prompts through adversarial training, achieving some improvement in both robustness and generalization. However, they primarily rely on singlegradient direction perturbations (e.g., PGD) to generate AEs, which lack diversity, resulting in limited improvement in adversarial robustness. To address these limitations, we propose an evolution-based region adversarial prompt tuning method called ER-APT, which combines gradient methods with genetic evolution to generate more diverse and challenging AEs. In each training iteration, we first generate AEs using traditional gradient-based methods. Subsequently, a genetic evolution mechanism incorporating selection, mutation, and crossover is applied to optimize the AEs, ensuring a broader and more aggressive perturbation distribution.The final evolved AEs are used for prompt tuning, achieving region-based adversarial optimization instead of conventional single-point adversarial prompt tuning. We also propose a dynamic loss weighting method to adjust prompt learning efficiency for accuracy and robustness. Experimental evaluations on various benchmark datasets demonstrate the superiority of our proposed method, outperforming stateof-the-art APT methods. The code is released at https://github.com/jiaxiaojunQAQ/ER-APT.

Scale-Invariant Adversarial Attack against Arbitrary-scale Super-resolution

The advent of local continuous image function (LIIF) has garnered significant attention for arbitrary-scale super-resolution (SR) techniques. However, while the vulnerabilities of fixed-scale SR have been assessed, the robustness of continuous representation-based arbitrary-scale SR against adversarial attacks remains an area warranting further exploration. The elaborately designed adversarial attacks for fixed-scale SR are scale-dependent, which will cause time-consuming and memory-consuming problems when applied to arbitrary-scale SR. To address this concern, we propose a simple yet effective ``scale-invariant'' SR adversarial attack method with good transferability, termed SIAGT. Specifically, we propose to construct resource-saving attacks by exploiting finite discrete points of continuous representation. In addition, we formulate a coordinate-dependent loss to enhance the cross-model transferability of the attack. The attack can significantly deteriorate the SR images while introducing imperceptible distortion to the targeted low-resolution (LR) images. Experiments carried out on three popular LIIF-based SR approaches and four classical SR datasets show remarkable attack performance and transferability of SIAGT.

PromptGuard: Soft Prompt-Guided Unsafe Content Moderation for Text-to-Image Models

Text-to-image (T2I) models have been shown to be vulnerable to misuse, particularly in generating not-safe-for-work (NSFW) content, raising serious ethical concerns. In this work, we present PromptGuard, a novel content moderation technique that draws inspiration from the system prompt mechanism in large language models (LLMs) for safety alignment. Unlike LLMs, T2I models lack a direct interface for enforcing behavioral guidelines. Our key idea is to optimize a safety soft prompt that functions as an implicit system prompt within the T2I model's textual embedding space. This universal soft prompt (P*) directly moderates NSFW inputs, enabling safe yet realistic image generation without altering the inference efficiency or requiring proxy models. Extensive experiments across three datasets demonstrate that PromptGuard effectively mitigates NSFW content generation while preserving high-quality benign outputs. PromptGuard achieves 7.8 times faster than prior content moderation methods, surpassing eight state-of-the-art defenses with an optimal unsafe ratio down to 5.84%.

Concept Guided Co-saliency Objection Detection

The task of co-saliency object detection (Co-SOD) seeks to identify common, salient objects across a collection of images by examining shared visual features. However, traditional Co-SOD methods often encounter limitations when faced with diverse object variations (e.g., different postures) and irrelevant background elements that introduce noise. To address these challenges, we propose ConceptCoSOD, a novel concept-guided approach that leverages text semantic information to enhance Co-SOD performance by guiding the model to focus on consistent object features. Through rethinking Co-SOD as an (image-text)-to-image task instead of an image-to-image task, ConceptCoSOD first captures shared semantic concepts within an image group and then uses them as guidance for precise object segmentation in complex scenarios. Experimental results on three benchmark datasets and six corruptions reveal that ConceptCoSOD significantly improves detection accuracy, especially in challenging settings with considerable background distractions and object variability.

What External Knowledge is Preferred by LLMs? Characterizing and Exploring Chain of Evidence in Imperfect Context

Incorporating external knowledge into large language models (LLMs) has emerged as a promising approach to mitigate outdated knowledge and hallucination in LLMs. However, external knowledge is often imperfect. In addition to useful knowledge, external knowledge is rich in irrelevant or misinformation in the context that can impair the reliability of LLM responses. This paper focuses on LLMs' preferred external knowledge in imperfect contexts when handling multi-hop QA. Inspired by criminal procedural law's Chain of Evidence (CoE), we characterize that knowledge preferred by LLMs should maintain both relevance to the question and mutual support among knowledge pieces. Accordingly, we propose an automated CoE discrimination approach and explore LLMs' preferences from their effectiveness, faithfulness and robustness, as well as CoE's usability in a naive Retrieval-Augmented Generation (RAG) case. The evaluation on five LLMs reveals that CoE enhances LLMs through more accurate generation, stronger answer faithfulness, better robustness against knowledge conflict, and improved performance in a popular RAG case.

Global Challenge for Safe and Secure LLMs Track 1

This paper introduces the Global Challenge for Safe and Secure Large Language Models (LLMs), a pioneering initiative organized by AI Singapore (AISG) and the CyberSG R&D Programme Office (CRPO) to foster the development of advanced defense mechanisms against automated jailbreaking attacks. With the increasing integration of LLMs in critical sectors such as healthcare, finance, and public administration, ensuring these models are resilient to adversarial attacks is vital for preventing misuse and upholding ethical standards. This competition focused on two distinct tracks designed to evaluate and enhance the robustness of LLM security frameworks. Track 1 tasked participants with developing automated methods to probe LLM vulnerabilities by eliciting undesirable responses, effectively testing the limits of existing safety protocols within LLMs. Participants were challenged to devise techniques that could bypass content safeguards across a diverse array of scenarios, from offensive language to misinformation and illegal activities. Through this process, Track 1 aimed to deepen the understanding of LLM vulnerabilities and provide insights for creating more resilient models.

Semantic-Aligned Adversarial Evolution Triangle for High-Transferability Vision-Language Attack

Vision-language pre-training (VLP) models excel at interpreting both images and text but remain vulnerable to multimodal adversarial examples (AEs). Advancing the generation of transferable AEs, which succeed across unseen models, is key to developing more robust and practical VLP models. Previous approaches augment image-text pairs to enhance diversity within the adversarial example generation process, aiming to improve transferability by expanding the contrast space of image-text features. However, these methods focus solely on diversity around the current AEs, yielding limited gains in transferability. To address this issue, we propose to increase the diversity of AEs by leveraging the intersection regions along the adversarial trajectory during optimization. Specifically, we propose sampling from adversarial evolution triangles composed of clean, historical, and current adversarial examples to enhance adversarial diversity. We provide a theoretical analysis to demonstrate the effectiveness of the proposed adversarial evolution triangle. Moreover, we find that redundant inactive dimensions can dominate similarity calculations, distorting feature matching and making AEs model-dependent with reduced transferability. Hence, we propose to generate AEs in the semantic image-text feature contrast space, which can project the original feature space into a semantic corpus subspace. The proposed semantic-aligned subspace can reduce the image feature redundancy, thereby improving adversarial transferability. Extensive experiments across different datasets and models demonstrate that the proposed method can effectively improve adversarial transferability and outperform state-of-the-art adversarial attack methods. The code is released at https://github.com/jiaxiaojunQAQ/SA-AET.