Zedian Shao

Zedian Shao, Hongbin Liu, Yuepeng Hu et al.

Multimodal large language models (MLLMs) have become the cornerstone of today's generative AI ecosystem, sparking intense competition among tech giants and startups. In particular, an MLLM generates a text response given a prompt consisting of an image and a question. While state-of-the-art MLLMs use safety filters and alignment techniques to refuse unsafe prompts, in this work, we introduce MLLM-Refusal, the first method that induces refusals for safe prompts. In particular, our MLLM-Refusal optimizes a nearly-imperceptible refusal perturbation and adds it to an image, causing target MLLMs to likely refuse a safe prompt containing the perturbed image and a safe question. Specifically, we formulate MLLM-Refusal as a constrained optimization problem and propose an algorithm to solve it. Our method offers competitive advantages for MLLM model providers by potentially disrupting user experiences of competing MLLMs, since competing MLLM's users will receive unexpected refusals when they unwittingly use these perturbed images in their prompts. We evaluate MLLM-Refusal on four MLLMs across four datasets, demonstrating its effectiveness in causing competing MLLMs to refuse safe prompts while not affecting non-competing MLLMs. Furthermore, we explore three potential countermeasures-adding Gaussian noise, DiffPure, and adversarial training. Our results show that though they can mitigate MLLM-Refusal's effectiveness, they also sacrifice the accuracy and/or efficiency of the competing MLLM. The code is available at https://github.com/Sadcardation/MLLM-Refusal.

Zedian Shao, Charles Fleming, Teodora Baluta

Large language models (LLMs) are often fine-tuned on uncurated text datasets that adversaries can poison. Existing poisoning attacks primarily rely on fixed trigger phrases that defenses such as outlier detection, clean-data regularization, or online monitoring can neutralize. In this paper, we propose a data poisoning method that teaches an LLM an information hiding scheme reliably and stealthily through semantic associations between shared knowledge such as facts or concepts and attacker-chosen phrases. The induced hiding scheme can encode and decode arbitrary malicious instructions, thus revealing a new and subtle poisoning-induced vulnerability: covert control attacks. We precisely characterize covert control attacks and evaluate them across $5$ LLMs, $3$ backdoor defenses, and $4$ prompt injection defenses. With a small poisoned fraction, covert control attacks outperform heuristic-based prompt injection attacks in average attack success rate by about $40\%$ relative to clean fine-tuned models. They also circumvent defenses based on detection and fine-tuning, maintaining up to $93\%$ attack success rate after backdoor defenses and up to $98\%$ after prompt injection defenses.

Yuqi Jia, Zedian Shao, Yupei Liu et al.

Large Language Models (LLMs) are vulnerable to prompt injection attacks, and several defenses have recently been proposed, often claiming to mitigate these attacks successfully. However, we argue that existing studies lack a principled approach to evaluating these defenses. In this paper, we argue the need to assess defenses across two critical dimensions: (1) effectiveness, measured against both existing and adaptive prompt injection attacks involving diverse target and injected prompts, and (2) general-purpose utility, ensuring that the defense does not compromise the foundational capabilities of the LLM. Our critical evaluation reveals that prior studies have not followed such a comprehensive evaluation methodology. When assessed using this principled approach, we show that existing defenses are not as successful as previously reported. This work provides a foundation for evaluating future defenses and guiding their development. Our code and data are available at: https://github.com/PIEval123/PIEval.

Zedian Shao, Hongbin Liu, Jaden Mu et al.

Prompt injection attack, where an attacker injects a prompt into the original one, aiming to make an Large Language Model (LLM) follow the injected prompt to perform an attacker-chosen task, represent a critical security threat. Existing attacks primarily focus on crafting these injections at inference time, treating the LLM itself as a static target. Our experiments show that these attacks achieve some success, but there is still significant room for improvement. In this work, we introduces a more foundational attack vector: poisoning the LLM's alignment process to amplify the success of future prompt injection attacks. Specifically, we propose PoisonedAlign, a method that strategically creates poisoned alignment samples to poison an LLM's alignment dataset. Our experiments across five LLMs and two alignment datasets show that when even a small fraction of the alignment data is poisoned, the resulting model becomes substantially more vulnerable to a wide range of prompt injection attacks. Crucially, this vulnerability is instilled while the LLM's performance on standard capability benchmarks remains largely unchanged, making the manipulation difficult to detect through automated, general-purpose performance evaluations. The code for implementing the attack is available at https://github.com/Sadcardation/PoisonedAlign.

Zedian Shao, Hongbin Liu, Yuepeng Hu et al.

Multi-modal large language models (MLLMs) have emerged as powerful tools for analyzing Internet-scale image data, offering significant benefits but also raising critical safety and societal concerns. In particular, open-weight MLLMs may be misused to extract sensitive information from personal images at scale, such as identities, locations, or other private details. In this work, we propose ImageProtector, a user-side method that proactively protects images before sharing by embedding a carefully crafted, nearly imperceptible perturbation that acts as a visual prompt injection attack on MLLMs. As a result, when an adversary analyzes a protected image with an MLLM, the MLLM is consistently induced to generate a refusal response such as "I'm sorry, I can't help with that request." We empirically demonstrate the effectiveness of ImageProtector across six MLLMs and four datasets. Additionally, we evaluate three potential countermeasures, Gaussian noise, DiffPure, and adversarial training, and show that while they partially mitigate the impact of ImageProtector, they simultaneously degrade model accuracy and/or efficiency. Our study focuses on the practically important setting of open-weight MLLMs and large-scale automated image analysis, and highlights both the promise and the limitations of perturbation-based privacy protection.

Zhongye Liu, Hongbin Liu, Yuepeng Hu et al.

Visual hallucination (VH) occurs when a multimodal large language model (MLLM) generates responses with incorrect visual details for prompts. Existing methods for generating VH test cases primarily rely on human annotations, typically in the form of triples: (image, question, answer). In this paper, we introduce VHExpansion, the first automated method for expanding VH test cases for MLLMs. Given an initial VH test case, VHExpansion automatically expands it by perturbing the question and answer through negation as well as modifying the image using both common and adversarial perturbations. Additionally, we propose a new evaluation metric, symmetric accuracy, which measures the proportion of correctly answered VH test-case pairs. Each pair consists of a test case and its negated counterpart. Our theoretical analysis shows that symmetric accuracy is an unbiased evaluation metric that remains unaffected by the imbalance of VH testing cases with varying answers when an MLLM is randomly guessing the answers, whereas traditional accuracy is prone to such imbalance. We apply VHExpansion to expand three VH datasets annotated manually and use these expanded datasets to benchmark seven MLLMs. Our evaluation shows that VHExpansion effectively identifies more VH test cases. Moreover, symmetric accuracy, being unbiased, leads to different conclusions about the vulnerability of MLLMs to VH compared to traditional accuracy metric. Finally, we show that fine-tuning MLLMs on the expanded VH dataset generated by VHExpansion mitigates VH more effectively than fine-tuning on the original, manually annotated dataset. Our code is available at: https://github.com/lycheeefish/VHExpansion.

Yuqi Jia, Yupei Liu, Zedian Shao et al.

Prompt injection attacks deceive a large language model into completing an attacker-specified task instead of its intended task by contaminating its input data with an injected prompt, which consists of injected instruction(s) and data. Localizing the injected prompt within contaminated data is crucial for post-attack forensic analysis and data recovery. Despite its growing importance, prompt injection localization remains largely unexplored. In this work, we bridge this gap by proposing PromptLocate, the first method for localizing injected prompts. PromptLocate comprises three steps: (1) splitting the contaminated data into semantically coherent segments, (2) identifying segments contaminated by injected instructions, and (3) pinpointing segments contaminated by injected data. We show PromptLocate accurately localizes injected prompts across eight existing and eight adaptive attacks.