This work demonstrates that LLM-based web navigation agents offer powerful automation capabilities but are vulnerable to Indirect Prompt Injection (IPI) attacks. We show that adversaries can embed universal adversarial triggers in webpage HTML to hijack agent behavior that utilizes the accessibility tree to parse HTML, causing unintended or malicious actions. Using the Greedy Coordinate Gradient (GCG) algorithm and a Browser Gym agent powered by Llama-3.1, our system demonstrates high success rates across real websites in both targeted and general attacks, including login credential exfiltration and forced ad clicks. Our empirical results highlight critical security risks and the need for stronger defenses as LLM-driven autonomous web agents become more widely adopted. The system software (https://github.com/sej2020/manipulating-web-agents) is released under the MIT License, with an accompanying publicly available demo website (http://lethaiq.github.io/attack-web-llm-agent).

Recent advancement in large-scale Artificial Intelligence (AI) models offering multimodal services have become foundational in AI systems, making them prime targets for model theft. Existing methods select Out-of-Distribution (OoD) data as backdoor watermarks and retrain the original model for copyright protection. However, existing methods are susceptible to malicious detection and forgery by adversaries, resulting in watermark evasion. In this work, we propose Model-\underline{ag}nostic Black-box Backdoor W\underline{ate}rmarking Framework (AGATE) to address stealthiness and robustness challenges in multimodal model copyright protection. Specifically, we propose an adversarial trigger generation method to generate stealthy adversarial triggers from ordinary dataset, providing visual fidelity while inducing semantic shifts. To alleviate the issue of anomaly detection among model outputs, we propose a post-transform module to correct the model output by narrowing the distance between adversarial trigger image embedding and text embedding. Subsequently, a two-phase watermark verification is proposed to judge whether the current model infringes by comparing the two results with and without the transform module. Consequently, we consistently outperform state-of-the-art methods across five datasets in the downstream tasks of multimodal image-text retrieval and image classification. Additionally, we validated the robustness of AGATE under two adversarial attack scenarios.

We investigate the robustness of reasoning models trained for step-by-step problem solving by introducing query-agnostic adversarial triggers - short, irrelevant text that, when appended to math problems, systematically mislead models to output incorrect answers without altering the problem's semantics. We propose CatAttack, an automated iterative attack pipeline for generating triggers on a weaker, less expensive proxy model (DeepSeek V3) and successfully transfer them to more advanced reasoning target models like DeepSeek R1 and DeepSeek R1-distilled-Qwen-32B, resulting in greater than 300% increase in the likelihood of the target model generating an incorrect answer. For example, appending, "Interesting fact: cats sleep most of their lives," to any math problem leads to more than doubling the chances of a model getting the answer wrong. Our findings highlight critical vulnerabilities in reasoning models, revealing that even state-of-the-art models remain susceptible to subtle adversarial inputs, raising security and reliability concerns. The CatAttack triggers dataset with model responses is available at https://huggingface.co/datasets/collinear-ai/cat-attack-adversarial-triggers.

Pre-trained models excel on NLI benchmarks like SNLI and MultiNLI, but their true language understanding remains uncertain. Models trained only on hypotheses and labels achieve high accuracy, indicating reliance on dataset biases and spurious correlations. To explore this issue, we applied the Universal Adversarial Attack to examine the model's vulnerabilities. Our analysis revealed substantial drops in accuracy for the entailment and neutral classes, whereas the contradiction class exhibited a smaller decline. Fine-tuning the model on an augmented dataset with adversarial examples restored its performance to near-baseline levels for both the standard and challenge sets. Our findings highlight the value of adversarial triggers in identifying spurious correlations and improving robustness while providing insights into the resilience of the contradiction class to adversarial attacks.

Large Language Models (LLMs) have demonstrated impressive capabilities in natural language tasks, but their safety and morality remain contentious due to their training on internet text corpora. To address these concerns, alignment techniques have been developed to improve the public usability and safety of LLMs. Yet, the potential for generating harmful content through these models seems to persist. This paper explores the concept of jailbreaking LLMs-reversing their alignment through adversarial triggers. Previous methods, such as soft embedding prompts, manually crafted prompts, and gradient-based automatic prompts, have had limited success on black-box models due to their requirements for model access and for producing a low variety of manually crafted prompts, making them susceptible to being blocked. This paper introduces a novel approach using reinforcement learning to optimize adversarial triggers, requiring only inference API access to the target model and a small surrogate model. Our method, which leverages a BERTScore-based reward function, enhances the transferability and effectiveness of adversarial triggers on new black-box models. We demonstrate that this approach improves the performance of adversarial triggers on a previously untested language model.

Recently, code-oriented large language models (Code LLMs) have been widely and successfully used to simplify and facilitate code programming. With these tools, developers can easily generate desired complete functional codes based on incomplete code and natural language prompts. However, a few pioneering works revealed that these Code LLMs are also vulnerable, e.g., against backdoor and adversarial attacks. The former could induce LLMs to respond to triggers to insert malicious code snippets by poisoning the training data or model parameters, while the latter can craft malicious adversarial input codes to reduce the quality of generated codes. However, both attack methods have underlying limitations: backdoor attacks rely on controlling the model training process, while adversarial attacks struggle with fulfilling specific malicious purposes. To inherit the advantages of both backdoor and adversarial attacks, this paper proposes a new attack paradigm, i.e., target-specific and adversarial prompt injection (TAPI), against Code LLMs. TAPI generates unreadable comments containing information about malicious instructions and hides them as triggers in the external source code. When users exploit Code LLMs to complete codes containing the trigger, the models will generate attacker-specified malicious code snippets at specific locations. We evaluate our TAPI attack on four representative LLMs under three representative malicious objectives and seven cases. The results show that our method is highly threatening (achieving an attack success rate enhancement of up to 89.3%) and stealthy (saving an average of 53.1% of tokens in the trigger design). In particular, we successfully attack some famous deployed code completion integrated applications, including CodeGeex and Github Copilot. This further confirms the realistic threat of our attack.

Recent work has developed optimization procedures to find token sequences, called adversarial triggers, which can elicit unsafe responses from aligned language models. These triggers are believed to be universally transferable, i.e., a trigger optimized on one model can jailbreak other models. In this paper, we concretely show that such adversarial triggers are not universal. We extensively investigate trigger transfer amongst 13 open models and observe inconsistent transfer. Our experiments further reveal a significant difference in robustness to adversarial triggers between models Aligned by Preference Optimization (APO) and models Aligned by Fine-Tuning (AFT). We find that APO models are extremely hard to jailbreak even when the trigger is optimized directly on the model. On the other hand, while AFT models may appear safe on the surface, exhibiting refusals to a range of unsafe instructions, we show that they are highly susceptible to adversarial triggers. Lastly, we observe that most triggers optimized on AFT models also generalize to new unsafe instructions from five diverse domains, further emphasizing their vulnerability. Overall, our work highlights the need for more comprehensive safety evaluations for aligned language models.

Prompt-based learning is a new language model training paradigm that adapts the Pre-trained Language Models (PLMs) to downstream tasks, which revitalizes the performance benchmarks across various natural language processing (NLP) tasks. Instead of using a fixed prompt template to fine-tune the model, some research demonstrates the effectiveness of searching for the prompt via optimization. Such prompt optimization process of prompt-based learning on PLMs also gives insight into generating adversarial prompts to mislead the model, raising concerns about the adversarial vulnerability of this paradigm. Recent studies have shown that universal adversarial triggers (UATs) can be generated to alter not only the predictions of the target PLMs but also the prediction of corresponding Prompt-based Fine-tuning Models (PFMs) under the prompt-based learning paradigm. However, UATs found in previous works are often unreadable tokens or characters and can be easily distinguished from natural texts with adaptive defenses. In this work, we consider the naturalness of the UATs and develop $\textit{LinkPrompt}$, an adversarial attack algorithm to generate UATs by a gradient-based beam search algorithm that not only effectively attacks the target PLMs and PFMs but also maintains the naturalness among the trigger tokens. Extensive results demonstrate the effectiveness of $\textit{LinkPrompt}$, as well as the transferability of UATs generated by $\textit{LinkPrompt}$ to open-sourced Large Language Model (LLM) Llama2 and API-accessed LLM GPT-3.5-turbo. The resource is available at $\href{https://github.com/SavannahXu79/LinkPrompt}{https://github.com/SavannahXu79/LinkPrompt}$.

The advent of Large Language Models (LLMs) has marked significant achievements in language processing and reasoning capabilities. Despite their advancements, LLMs face vulnerabilities to data poisoning attacks, where adversaries insert backdoor triggers into training data to manipulate outputs for malicious purposes. This work further identifies additional security risks in LLMs by designing a new data poisoning attack tailored to exploit the instruction tuning process. We propose a novel gradient-guided backdoor trigger learning approach to identify adversarial triggers efficiently, ensuring an evasion of detection by conventional defenses while maintaining content integrity. Through experimental validation across various LLMs and tasks, our strategy demonstrates a high success rate in compromising model outputs; poisoning only 1\% of 4,000 instruction tuning samples leads to a Performance Drop Rate (PDR) of around 80\%. Our work highlights the need for stronger defenses against data poisoning attack, offering insights into safeguarding LLMs against these more sophisticated attacks. The source code can be found on this GitHub repository: https://github.com/RookieZxy/GBTL/blob/main/README.md.

Neural text ranking models have witnessed significant advancement and are increasingly being deployed in practice. Unfortunately, they also inherit adversarial vulnerabilities of general neural models, which have been detected but remain underexplored by prior studies. Moreover, the inherit adversarial vulnerabilities might be leveraged by blackhat SEO to defeat better-protected search engines. In this study, we propose an imitation adversarial attack on black-box neural passage ranking models. We first show that the target passage ranking model can be transparentized and imitated by enumerating critical queries/candidates and then train a ranking imitation model. Leveraging the ranking imitation model, we can elaborately manipulate the ranking results and transfer the manipulation attack to the target ranking model. For this purpose, we propose an innovative gradient-based attack method, empowered by the pairwise objective function, to generate adversarial triggers, which causes premeditated disorderliness with very few tokens. To equip the trigger camouflages, we add the next sentence prediction loss and the language model fluency constraint to the objective function. Experimental results on passage ranking demonstrate the effectiveness of the ranking imitation attack model and adversarial triggers against various SOTA neural ranking models. Furthermore, various mitigation analyses and human evaluation show the effectiveness of camouflages when facing potential mitigation approaches. To motivate other scholars to further investigate this novel and important problem, we make the experiment data and code publicly available.