Manipulating Multimodal Agents via Cross-Modal Prompt Injection

The emergence of multimodal large language models has redefined the agent paradigm by integrating language and vision modalities with external data sources, enabling agents to better interpret human instructions and execute increasingly complex tasks. However, in this work, we identify a critical yet previously overlooked security vulnerability in multimodal agents: cross-modal prompt injection attacks. To exploit this vulnerability, we propose CrossInject, a novel attack framework in which attackers embed adversarial perturbations across multiple modalities to align with target malicious content, allowing external instructions to hijack the agent's decision-making process and execute unauthorized tasks. Our approach consists of two key components. First, we introduce Visual Latent Alignment, where we optimize adversarial features to the malicious instructions in the visual embedding space based on a text-to-image generative model, ensuring that adversarial images subtly encode cues for malicious task execution. Subsequently, we present Textual Guidance Enhancement, where a large language model is leveraged to infer the black-box defensive system prompt through adversarial meta prompting and generate an malicious textual command that steers the agent's output toward better compliance with attackers' requests. Extensive experiments demonstrate that our method outperforms existing injection attacks, achieving at least a +26.4% increase in attack success rates across diverse tasks. Furthermore, we validate our attack's effectiveness in real-world multimodal autonomous agents, highlighting its potential implications for safety-critical applications.

Secure Transfer Learning: Training Clean Models Against Backdoor in (Both) Pre-trained Encoders and Downstream Datasets

Transfer learning from pre-trained encoders has become essential in modern machine learning, enabling efficient model adaptation across diverse tasks. However, this combination of pre-training and downstream adaptation creates an expanded attack surface, exposing models to sophisticated backdoor embeddings at both the encoder and dataset levels--an area often overlooked in prior research. Additionally, the limited computational resources typically available to users of pre-trained encoders constrain the effectiveness of generic backdoor defenses compared to end-to-end training from scratch. In this work, we investigate how to mitigate potential backdoor risks in resource-constrained transfer learning scenarios. Specifically, we conduct an exhaustive analysis of existing defense strategies, revealing that many follow a reactive workflow based on assumptions that do not scale to unknown threats, novel attack types, or different training paradigms. In response, we introduce a proactive mindset focused on identifying clean elements and propose the Trusted Core (T-Core) Bootstrapping framework, which emphasizes the importance of pinpointing trustworthy data and neurons to enhance model security. Our empirical evaluations demonstrate the effectiveness and superiority of T-Core, specifically assessing 5 encoder poisoning attacks, 7 dataset poisoning attacks, and 14 baseline defenses across five benchmark datasets, addressing four scenarios of 3 potential backdoor threats.

Exploring Gradient-Guided Masked Language Model to Detect Textual Adversarial Attacks

Textual adversarial examples pose serious threats to the reliability of natural language processing systems. Recent studies suggest that adversarial examples tend to deviate from the underlying manifold of normal texts, whereas pre-trained masked language models can approximate the manifold of normal data. These findings inspire the exploration of masked language models for detecting textual adversarial attacks. We first introduce Masked Language Model-based Detection (MLMD), leveraging the mask and unmask operations of the masked language modeling (MLM) objective to induce the difference in manifold changes between normal and adversarial texts. Although MLMD achieves competitive detection performance, its exhaustive one-by-one masking strategy introduces significant computational overhead. Our posterior analysis reveals that a significant number of non-keywords in the input are not important for detection but consume resources. Building on this, we introduce Gradient-guided MLMD (GradMLMD), which leverages gradient information to identify and skip non-keywords during detection, significantly reducing resource consumption without compromising detection performance.

Test-Time Backdoor Detection for Object Detection Models

Object detection models are vulnerable to backdoor attacks, where attackers poison a small subset of training samples by embedding a predefined trigger to manipulate prediction. Detecting poisoned samples (i.e., those containing triggers) at test time can prevent backdoor activation. However, unlike image classification tasks, the unique characteristics of object detection -- particularly its output of numerous objects -- pose fresh challenges for backdoor detection. The complex attack effects (e.g., "ghost" object emergence or "vanishing" object) further render current defenses fundamentally inadequate. To this end, we design TRAnsformation Consistency Evaluation (TRACE), a brand-new method for detecting poisoned samples at test time in object detection. Our journey begins with two intriguing observations: (1) poisoned samples exhibit significantly more consistent detection results than clean ones across varied backgrounds. (2) clean samples show higher detection consistency when introduced to different focal information. Based on these phenomena, TRACE applies foreground and background transformations to each test sample, then assesses transformation consistency by calculating the variance in objects confidences. TRACE achieves black-box, universal backdoor detection, with extensive experiments showing a 30% improvement in AUROC over state-of-the-art defenses and resistance to adaptive attacks.

Improving Generalization of Universal Adversarial Perturbation via Dynamic Maximin Optimization

Deep neural networks (DNNs) are susceptible to universal adversarial perturbations (UAPs). These perturbations are meticulously designed to fool the target model universally across all sample classes. Unlike instance-specific adversarial examples (AEs), generating UAPs is more complex because they must be generalized across a wide range of data samples and models. Our research reveals that existing universal attack methods, which optimize UAPs using DNNs with static model parameter snapshots, do not fully leverage the potential of DNNs to generate more effective UAPs. Rather than optimizing UAPs against static DNN models with a fixed training set, we suggest using dynamic model-data pairs to generate UAPs. In particular, we introduce a dynamic maximin optimization strategy, aiming to optimize the UAP across a variety of optimal model-data pairs. We term this approach DM-UAP. DM-UAP utilizes an iterative max-min-min optimization framework that refines the model-data pairs, coupled with a curriculum UAP learning algorithm to examine the combined space of model parameters and data thoroughly. Comprehensive experiments on the ImageNet dataset demonstrate that the proposed DM-UAP markedly enhances both cross-sample universality and cross-model transferability of UAPs. Using only 500 samples for UAP generation, DM-UAP outperforms the state-of-the-art approach with an average increase in fooling ratio of 12.108%.

ViDTA: Enhanced Drug-Target Affinity Prediction via Virtual Graph Nodes and Attention-based Feature Fusion

Drug-target interaction is fundamental in understanding how drugs affect biological systems, and accurately predicting drug-target affinity (DTA) is vital for drug discovery. Recently, deep learning methods have emerged as a significant approach for estimating the binding strength between drugs and target proteins. However, existing methods simply utilize the drug's local information from molecular topology rather than global information. Additionally, the features of drugs and proteins are usually fused with a simple concatenation operation, limiting their effectiveness. To address these challenges, we proposed ViDTA, an enhanced DTA prediction framework. We introduce virtual nodes into the Graph Neural Network (GNN)-based drug feature extraction network, which acts as a global memory to exchange messages more efficiently. By incorporating virtual graph nodes, we seamlessly integrate local and global features of drug molecular structures, expanding the GNN's receptive field. Additionally, we propose an attention-based linear feature fusion network for better capturing the interaction information between drugs and proteins. Experimental results evaluated on various benchmarks including Davis, Metz, and KIBA demonstrate that our proposed ViDTA outperforms the state-of-the-art baselines.

Breaking Barriers in Physical-World Adversarial Examples: Improving Robustness and Transferability via Robust Feature

As deep neural networks (DNNs) are widely applied in the physical world, many researches are focusing on physical-world adversarial examples (PAEs), which introduce perturbations to inputs and cause the model's incorrect outputs. However, existing PAEs face two challenges: unsatisfactory attack performance (i.e., poor transferability and insufficient robustness to environment conditions), and difficulty in balancing attack effectiveness with stealthiness, where better attack effectiveness often makes PAEs more perceptible. In this paper, we explore a novel perturbation-based method to overcome the challenges. For the first challenge, we introduce a strategy Deceptive RF injection based on robust features (RFs) that are predictive, robust to perturbations, and consistent across different models. Specifically, it improves the transferability and robustness of PAEs by covering RFs of other classes onto the predictive features in clean images. For the second challenge, we introduce another strategy Adversarial Semantic Pattern Minimization, which removes most perturbations and retains only essential adversarial patterns in AEsBased on the two strategies, we design our method Robust Feature Coverage Attack (RFCoA), comprising Robust Feature Disentanglement and Adversarial Feature Fusion. In the first stage, we extract target class RFs in feature space. In the second stage, we use attention-based feature fusion to overlay these RFs onto predictive features of clean images and remove unnecessary perturbations. Experiments show our method's superior transferability, robustness, and stealthiness compared to existing state-of-the-art methods. Additionally, our method's effectiveness can extend to Large Vision-Language Models (LVLMs), indicating its potential applicability to more complex tasks.

NumbOD: A Spatial-Frequency Fusion Attack Against Object Detectors

With the advancement of deep learning, object detectors (ODs) with various architectures have achieved significant success in complex scenarios like autonomous driving. Previous adversarial attacks against ODs have been focused on designing customized attacks targeting their specific structures (e.g., NMS and RPN), yielding some results but simultaneously constraining their scalability. Moreover, most efforts against ODs stem from image-level attacks originally designed for classification tasks, resulting in redundant computations and disturbances in object-irrelevant areas (e.g., background). Consequently, how to design a model-agnostic efficient attack to comprehensively evaluate the vulnerabilities of ODs remains challenging and unresolved. In this paper, we propose NumbOD, a brand-new spatial-frequency fusion attack against various ODs, aimed at disrupting object detection within images. We directly leverage the features output by the OD without relying on its internal structures to craft adversarial examples. Specifically, we first design a dual-track attack target selection strategy to select high-quality bounding boxes from OD outputs for targeting. Subsequently, we employ directional perturbations to shift and compress predicted boxes and change classification results to deceive ODs. Additionally, we focus on manipulating the high-frequency components of images to confuse ODs' attention on critical objects, thereby enhancing the attack efficiency. Our extensive experiments on nine ODs and two datasets show that NumbOD achieves powerful attack performance and high stealthiness.

PB-UAP: Hybrid Universal Adversarial Attack For Image Segmentation

With the rapid advancement of deep learning, the model robustness has become a significant research hotspot, \ie, adversarial attacks on deep neural networks. Existing works primarily focus on image classification tasks, aiming to alter the model's predicted labels. Due to the output complexity and deeper network architectures, research on adversarial examples for segmentation models is still limited, particularly for universal adversarial perturbations. In this paper, we propose a novel universal adversarial attack method designed for segmentation models, which includes dual feature separation and low-frequency scattering modules. The two modules guide the training of adversarial examples in the pixel and frequency space, respectively. Experiments demonstrate that our method achieves high attack success rates surpassing the state-of-the-art methods, and exhibits strong transferability across different models.

Visual Adversarial Attack on Vision-Language Models for Autonomous Driving

Vision-language models (VLMs) have significantly advanced autonomous driving (AD) by enhancing reasoning capabilities. However, these models remain highly vulnerable to adversarial attacks. While existing research has primarily focused on general VLM attacks, the development of attacks tailored to the safety-critical AD context has been largely overlooked. In this paper, we take the first step toward designing adversarial attacks specifically targeting VLMs in AD, exposing the substantial risks these attacks pose within this critical domain. We identify two unique challenges for effective adversarial attacks on AD VLMs: the variability of textual instructions and the time-series nature of visual scenarios. To this end, we propose ADvLM, the first visual adversarial attack framework specifically designed for VLMs in AD. Our framework introduces Semantic-Invariant Induction, which uses a large language model to create a diverse prompt library of textual instructions with consistent semantic content, guided by semantic entropy. Building on this, we introduce Scenario-Associated Enhancement, an approach where attention mechanisms select key frames and perspectives within driving scenarios to optimize adversarial perturbations that generalize across the entire scenario. Extensive experiments on several AD VLMs over multiple benchmarks show that ADvLM achieves state-of-the-art attack effectiveness. Moreover, real-world attack studies further validate its applicability and potential in practice.