arXiv subjects

Jin Song Dong

Publications and source records attributed to Jin Song Dong.

At least 55 records · Page 3Linked to original sources

arXiv2025

Bones of Contention: Exploring Query-Efficient Attacks against Skeleton Recognition Systems

Skeleton action recognition models have secured more attention than video-based ones in various applications due to privacy preservation and lower storage requirements. Skeleton data are typically transmitted to cloud servers for action recognition, with results returned to clients via Apps/APIs. However, the vulnerability of skeletal models against adversarial perturbations gradually reveals the unreliability of these systems. Existing black-box attacks all operate in a decision-based manner, resulting in numerous queries that hinder efficiency and feasibility in real-world applications. Moreover, all attacks off the shelf focus on only restricted perturbations, while ignoring model weaknesses when encountered with non-semantic perturbations. In this paper, we propose two query-effIcient Skeletal Adversarial AttaCks, ISAAC-K and ISAAC-N. As a black-box attack, ISAAC-K utilizes Grad-CAM in a surrogate model to extract key joints where minor sparse perturbations are then added to fool the classifier. To guarantee natural adversarial motions, we introduce constraints of both bone length and temporal consistency. ISAAC-K finds stronger adversarial examples on the $\ell_\infty$ norm, which can encompass those on other norms. Exhaustive experiments substantiate that ISAAC-K can uplift the attack efficiency of the perturbations under 10 skeletal models. Additionally, as a byproduct, ISAAC-N fools the classifier by replacing skeletons unrelated to the action. We surprisingly find that skeletal models are vulnerable to large perturbations where the part-wise non-semantic joints are just replaced, leading to a query-free no-box attack without any prior knowledge. Based on that, four adaptive defenses are eventually proposed to improve the robustness of skeleton recognition models.

cs.CR↗

arXiv2025

RvLLM: LLM Runtime Verification with Domain Knowledge

Large language models (LLMs) have emerged as a dominant AI paradigm due to their exceptional text understanding and generation capabilities. However, their tendency to generate inconsistent or erroneous outputs challenges their reliability, especially in high-stakes domains requiring accuracy and trustworthiness. Existing research primarily focuses on detecting and mitigating model misbehavior in general-purpose scenarios, often overlooking the potential of integrating domain-specific knowledge. In this work, we advance misbehavior detection by incorporating domain knowledge. The core idea is to design a general specification language that enables domain experts to customize domain-specific predicates in a lightweight and intuitive manner, supporting later runtime verification of LLM outputs. To achieve this, we design a novel specification language, ESL, and introduce a runtime verification framework, RvLLM, to validate LLM output against domain-specific constraints defined in ESL. We evaluate RvLLM on three representative tasks: violation detection against Singapore Rapid Transit Systems Act, numerical comparison, and inequality solving. Experimental results demonstrate that RvLLM effectively detects erroneous outputs across various LLMs in a lightweight and flexible manner. The results reveal that despite their impressive capabilities, LLMs remain prone to low-level errors due to limited interpretability and a lack of formal guarantees during inference, and our framework offers a potential long-term solution by leveraging expert domain knowledge to rigorously and efficiently verify LLM outputs.

cs.AI↗

arXiv2025

DRIP: Defending Prompt Injection via Token-wise Representation Editing and Residual Instruction Fusion

Large language models (LLMs) are increasingly integrated into IT infrastructures, where they process user data according to predefined instructions. However, conventional LLMs remain vulnerable to prompt injection, where malicious users inject directive tokens into the data to subvert model behavior. Existing defenses train LLMs to semantically separate data and instruction tokens, but still struggle to (1) balance utility and security and (2) prevent instruction-like semantics in the data from overriding the intended instructions. We propose DRIP, which (1) precisely removes instruction semantics from tokens in the data section while preserving their data semantics, and (2) robustly preserves the effect of the intended instruction even under strong adversarial content. To "de-instructionalize" data tokens, DRIP introduces a data curation and training paradigm with a lightweight representation-editing module that edits embeddings of instruction-like tokens in the data section, enhancing security without harming utility. To ensure non-overwritability of instructions, DRIP adds a minimal residual module that reduces the ability of adversarial data to overwrite the original instruction. We evaluate DRIP on LLaMA 8B and Mistral 7B against StruQ, SecAlign, ISE, and PFT on three prompt-injection benchmarks (SEP, AlpacaFarm, and InjecAgent). DRIP improves role-separation score by 12-49\%, reduces attack success rate by over 66\% under adaptive attacks, and matches the utility of the undefended model, establishing a new state of the art for prompt-injection robustness.

cs.CR↗

arXiv2025

Few-Shot Precise Event Spotting via Unified Multi-Entity Graph and Distillation

Precise event spotting (PES) aims to recognize fine-grained events at exact moments and has become a key component of sports analytics. This task is particularly challenging due to rapid succession, motion blur, and subtle visual differences. Consequently, most existing methods rely on domain-specific, end-to-end training with large labeled datasets and often struggle in few-shot conditions due to their dependence on pixel- or pose-based inputs alone. However, obtaining large labeled datasets is practically hard. We propose a Unified Multi-Entity Graph Network (UMEG-Net) for few-shot PES. UMEG-Net integrates human skeletons and sport-specific object keypoints into a unified graph and features an efficient spatio-temporal extraction module based on advanced GCN and multi-scale temporal shift. To further enhance performance, we employ multimodal distillation to transfer knowledge from keypoint-based graphs to visual representations. Our approach achieves robust performance with limited labeled data and significantly outperforms baseline models in few-shot settings, providing a scalable and effective solution for few-shot PES. Code is publicly available at https://github.com/LZYAndy/UMEG-Net.

cs.CV↗

arXiv2025

Failures to Surface Harmful Contents in Video Large Language Models

Video Large Language Models (VideoLLMs) are increasingly deployed on numerous critical applications, where users rely on auto-generated summaries while casually skimming the video stream. We show that this interaction hides a critical safety gap: if harmful content is embedded in a video, either as full-frame inserts or as small corner patches, state-of-the-art VideoLLMs rarely mention the harmful content in the output, despite its clear visibility to human viewers. A root-cause analysis reveals three compounding design flaws: (1) insufficient temporal coverage resulting from the sparse, uniformly spaced frame sampling used by most leading VideoLLMs, (2) spatial information loss introduced by aggressive token downsampling within sampled frames, and (3) encoder-decoder disconnection, whereby visual cues are only weakly utilized during text generation. Leveraging these insights, we craft three zero-query black-box attacks, aligning with these flaws in the processing pipeline. Our large-scale evaluation across five leading VideoLLMs shows that the harmfulness omission rate exceeds 90% in most cases. Even when harmful content is clearly present in all frames, these models consistently fail to identify it. These results underscore a fundamental vulnerability in current VideoLLMs' designs and highlight the urgent need for sampling strategies, token compression, and decoding mechanisms that guarantee semantic coverage rather than speed alone.

cs.MM↗

arXiv2025

RulePilot: An LLM-Powered Agent for Security Rule Generation

The real-time demand for system security leads to the detection rules becoming an integral part of the intrusion detection life-cycle. Rule-based detection often identifies malicious logs based on the predefined grammar logic, requiring experts with deep domain knowledge for rule generation. Therefore, automation of rule generation can result in significant time savings and ease the burden of rule-related tasks on security engineers. In this paper, we propose RulePilot, which mimics human expertise via LLM-based agent for addressing rule-related challenges like rule creation or conversion. Using RulePilot, the security analysts do not need to write down the rules following the grammar, instead, they can just provide the annotations such as the natural-language-based descriptions of a rule, our RulePilot can automatically generate the detection rules without more intervention. RulePilot is equipped with the intermediate representation (IR), which abstracts the complexity of config rules into structured, standardized formats, allowing LLMs to focus on generation rules in a more manageable and consistent way. We present a comprehensive evaluation of RulePilot in terms of textual similarity and execution success abilities, showcasing RulePilot can generate high-fidelity rules, outperforming the baseline models by up to 107.4% in textual similarity to ground truths and achieving better detection accuracy in real-world execution tests. We perform a case study from our industry collaborators in Singapore, showcasing that RulePilot significantly help junior analysts/general users in the rule creation process.

cs.CR↗

arXiv2025

On the Mistaken Assumption of Interchangeable Deep Reinforcement Learning Implementations

Deep Reinforcement Learning (DRL) is a paradigm of artificial intelligence where an agent uses a neural network to learn which actions to take in a given environment. DRL has recently gained traction from being able to solve complex environments like driving simulators, 3D robotic control, and multiplayer-online-battle-arena video games. Numerous implementations of the state-of-the-art algorithms responsible for training these agents, like the Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) algorithms, currently exist. However, studies make the mistake of assuming implementations of the same algorithm to be consistent and thus, interchangeable. In this paper, through a differential testing lens, we present the results of studying the extent of implementation inconsistencies, their effect on the implementations' performance, as well as their impact on the conclusions of prior studies under the assumption of interchangeable implementations. The outcomes of our differential tests showed significant discrepancies between the tested algorithm implementations, indicating that they are not interchangeable. In particular, out of the five PPO implementations tested on 56 games, three implementations achieved superhuman performance for 50% of their total trials while the other two implementations only achieved superhuman performance for less than 15% of their total trials. As part of a meticulous manual analysis of the implementations' source code, we analyzed implementation discrepancies and determined that code-level inconsistencies primarily caused these discrepancies. Lastly, we replicated a study and showed that this assumption of implementation interchangeability was sufficient to flip experiment outcomes. Therefore, this calls for a shift in how implementations are being used.

cs.SE↗

arXiv2025

ALMGuard: Safety Shortcuts and Where to Find Them as Guardrails for Audio-Language Models

Recent advances in Audio-Language Models (ALMs) have significantly improved multimodal understanding capabilities. However, the introduction of the audio modality also brings new and unique vulnerability vectors. Previous studies have proposed jailbreak attacks that specifically target ALMs, revealing that defenses directly transferred from traditional audio adversarial attacks or text-based Large Language Model (LLM) jailbreaks are largely ineffective against these ALM-specific threats. To address this issue, we propose ALMGuard, the first defense framework tailored to ALMs. Based on the assumption that safety-aligned shortcuts naturally exist in ALMs, we design a method to identify universal Shortcut Activation Perturbations (SAPs) that serve as triggers that activate the safety shortcuts to safeguard ALMs at inference time. To better sift out effective triggers while preserving the model's utility on benign tasks, we further propose Mel-Gradient Sparse Mask (M-GSM), which restricts perturbations to Mel-frequency bins that are sensitive to jailbreaks but insensitive to speech understanding. Both theoretical analyses and empirical results demonstrate the robustness of our method against both seen and unseen attacks. Overall, \MethodName reduces the average success rate of advanced ALM-specific jailbreak attacks to 4.6% across four models, while maintaining comparable utility on benign benchmarks, establishing it as the new state of the art. Our code and data are available at https://github.com/WeifeiJin/ALMGuard.

cs.SD↗

arXiv2025

Neural Surveillance: Live-Update Visualization of Latent Training Dynamics

Monitoring the inner state of deep neural networks is essential for auditing the learning process and enabling timely interventions. While conventional metrics like validation loss offer a surface-level view of performance, the evolution of a model's hidden representations provides a deeper, complementary window into its internal dynamics. However, the literature lacks a real-time tool to monitor these crucial internal states. To address this, we introduce SentryCam, a live-update visualization framework that tracks the progression of hidden representations throughout training. SentryCam produces high-fidelity visualizations of the evolving representation space with minimal latency, serving as a powerful dashboard for understanding how a model learns. We quantitatively validate the faithfulness of SentryCam's visualizations across diverse datasets and architectures (ResNet, ViT). Furthermore, we demonstrate SentryCam's practical utility for model auditing through a case study on training instability. We designed an automated auditing system with geometry-based alerts that successfully identified impending model failure up to 7 epochs earlier than was evident from the validation loss curve. SentryCam's flexible framework is easily adaptable, supporting both the exploratory analysis and proactive auditing essential for robust model development. The code is available at https://github.com/xianglinyang/SentryCam.

cs.LG↗

arXiv2025

APREBot: Active Perception System for Reflexive Evasion Robot

Reliable onboard perception is critical for quadruped robots navigating dynamic environments, where obstacles can emerge from any direction under strict reaction-time constraints. Single-sensor systems face inherent limitations: LiDAR provides omnidirectional coverage but lacks rich texture information, while cameras capture high-resolution detail but suffer from restricted field of view. We introduce APREBot (Active Perception System for Reflexive Evasion Robot), a novel framework that integrates reflexive evasion with active hierarchical perception. APREBot strategically combines LiDAR-based omnidirectional scanning with camera-based active focusing, achieving comprehensive environmental awareness essential for agile obstacle avoidance in quadruped robots. We validate APREBot through extensive sim-to-real experiments on a quadruped platform, evaluating diverse obstacle types, trajectories, and approach directions. Our results demonstrate substantial improvements over state-of-the-art baselines in both safety metrics and operational efficiency, highlighting APREBot's potential for dependable autonomy in safety-critical scenarios. Videos are available at https://sites.google.com/view/aprebot/

cs.RO↗

arXiv2025

PAT-Agent: Autoformalization for Model Checking

Recent advances in large language models (LLMs) offer promising potential for automating formal methods. However, applying them to formal verification remains challenging due to the complexity of specification languages, the risk of hallucinated output, and the semantic gap between natural language and formal logic. We introduce PAT-Agent, an end-to-end framework for natural language autoformalization and formal model repair that combines the generative capabilities of LLMs with the rigor of formal verification to automate the construction of verifiable formal models. In PAT-Agent, a Planning LLM first extracts key modeling elements and generates a detailed plan using semantic prompts, which then guides a Code Generation LLM to synthesize syntactically correct and semantically faithful formal models. The resulting code is verified using the Process Analysis Toolkit (PAT) model checker against user-specified properties, and when discrepancies occur, a Repair Loop is triggered to iteratively correct the model using counterexamples. To improve flexibility, we built a web-based interface that enables users, particularly non-FM-experts, to describe, customize, and verify system behaviors through user-LLM interactions. Experimental results on 40 systems show that PAT-Agent consistently outperforms baselines, achieving high verification success with superior efficiency. The ablation studies confirm the importance of both planning and repair components, and the user study demonstrates that our interface is accessible and supports effective formal modeling, even for users with limited formal methods experience.

cs.SE↗

arXiv2025

Defending LVLMs Against Vision Attacks through Partial-Perception Supervision

Recent studies have raised significant concerns regarding the vulnerability of Large Vision Language Models (LVLMs) to maliciously injected or perturbed input images, which can mislead their responses. Existing defense methods show that such vision attacks are sensitive to image modifications especially cropping, using majority voting across responses of modified images as corrected responses. However, these modifications often result in partial images and distort the semantics, which reduces response quality on clean images after voting. Instead of directly using responses from partial images for voting, we investigate using them to supervise the LVLM's responses to the original images. We propose a black-box, training-free method called DPS (Defense through Partial-Perception Supervision). In this approach, the model is prompted using the responses generated by a model that perceives only a partial image. With DPS, the model can adjust its response based on partial image understanding when under attack, while confidently maintaining its original response for clean input. Our findings show that the weak model can supervise the strong model: when faced with an attacked input, the strong model becomes less confident and adjusts its response based on the weak model's partial understanding, effectively defending against the attack. With clean input, it confidently maintains its original response. Empirical experiments show our method outperforms the baseline, cutting the average attack success rate by 76.3% across six datasets on three popular models.

cs.CV↗

arXiv2025

Towards Powerful and Practical Patch Attacks for 2D Object Detection in Autonomous Driving

Learning-based autonomous driving systems remain critically vulnerable to adversarial patches, posing serious safety and security risks in their real-world deployment. Black-box attacks, notable for their high attack success rate without model knowledge, are especially concerning, with their transferability extensively studied to reduce computational costs compared to query-based attacks. Previous transferability-based black-box attacks typically adopt mean Average Precision (mAP) as the evaluation metric and design training loss accordingly. However, due to the presence of multiple detected bounding boxes and the relatively lenient Intersection over Union (IoU) thresholds, the attack effectiveness of these approaches is often overestimated, resulting in reduced success rates in practical attacking scenarios. Furthermore, patches trained on low-resolution data often fail to maintain effectiveness on high-resolution images, limiting their transferability to autonomous driving datasets. To fill this gap, we propose P$^3$A, a Powerful and Practical Patch Attack framework for 2D object detection in autonomous driving, specifically optimized for high-resolution datasets. First, we introduce a novel metric, Practical Attack Success Rate (PASR), to more accurately quantify attack effectiveness with greater relevance for pedestrian safety. Second, we present a tailored Localization-Confidence Suppression Loss (LCSL) to improve attack transferability under PASR. Finally, to maintain the transferability for high-resolution datasets, we further incorporate the Probabilistic Scale-Preserving Padding (PSPP) into the patch attack pipeline as a data preprocessing step. Extensive experiments show that P$^3$A outperforms state-of-the-art attacks on unseen models and unseen high-resolution datasets, both under the proposed practical IoU-based evaluation metric and the previous mAP-based metrics.

cs.CV↗

arXiv2025

REBot: Reflexive Evasion Robot for Instantaneous Dynamic Obstacle Avoidance

Dynamic obstacle avoidance (DOA) is critical for quadrupedal robots operating in environments with moving obstacles or humans. Existing approaches typically rely on navigation-based trajectory replanning, which assumes sufficient reaction time and leading to fails when obstacles approach rapidly. In such scenarios, quadrupedal robots require reflexive evasion capabilities to perform instantaneous, low-latency maneuvers. This paper introduces Reflexive Evasion Robot (REBot), a control framework that enables quadrupedal robots to achieve real-time reflexive obstacle avoidance. REBot integrates an avoidance policy and a recovery policy within a finite-state machine. With carefully designed learning curricula and by incorporating regularization and adaptive rewards, REBot achieves robust evasion and rapid stabilization in instantaneous DOA tasks. We validate REBot through extensive simulations and real-world experiments, demonstrating notable improvements in avoidance success rates, energy efficiency, and robustness to fast-moving obstacles. Videos and appendix are available on https://rebot-2025.github.io/.

cs.RO↗

arXiv2025

Seeing It Before It Happens: In-Generation NSFW Detection for Diffusion-Based Text-to-Image Models

Diffusion-based text-to-image (T2I) models enable high-quality image generation but also pose significant risks of misuse, particularly in producing not-safe-for-work (NSFW) content. While prior detection methods have focused on filtering prompts before generation or moderating images afterward, the in-generation phase of diffusion models remains largely unexplored for NSFW detection. In this paper, we introduce In-Generation Detection (IGD), a simple yet effective approach that leverages the predicted noise during the diffusion process as an internal signal to identify NSFW content. This approach is motivated by preliminary findings suggesting that the predicted noise may capture semantic cues that differentiate NSFW from benign prompts, even when the prompts are adversarially crafted. Experiments conducted on seven NSFW categories show that IGD achieves an average detection accuracy of 91.32% over naive and adversarial NSFW prompts, outperforming seven baseline methods.

cs.CV↗

arXiv2025

GUI Test Migration via Abstraction and Concretization

GUI test migration aims to produce test cases with events and assertions to test specific functionalities of a target app. Existing migration approaches typically focus on the widget-mapping paradigm that maps widgets from source apps to target apps. However, since different apps may implement the same functionality in different ways, direct mapping may result in incomplete or buggy test cases, thus significantly impacting the effectiveness of testing target functionality and the practical applicability of migration approaches. In this paper, we propose a new migration paradigm (i.e., the abstraction-concretization paradigm) that first abstracts the test logic for the target functionality and then utilizes this logic to generate the concrete GUI test case. Furthermore, we introduce MACdroid, the first approach that migrates GUI test cases based on this paradigm. Specifically, we propose an abstraction technique that utilizes source test cases from source apps targeting the same functionality to extract a general test logic for that functionality. Then, we propose a concretization technique that utilizes the general test logic to guide an LLM in generating the corresponding GUI test case (including events and assertions) for the target app. We evaluate MACdroid on two widely-used datasets (including 31 apps, 34 functionalities, and 123 test cases). On the FrUITeR dataset, the test cases generated by MACdroid successfully test 64% of the target functionalities, improving the baselines by 191%. On the Lin dataset, MACdroid successfully tests 75% of the target functionalities, outperforming the baselines by 42%. These results underscore the effectiveness of MACdroid in GUI test migration.

cs.SE↗

arXiv2025

PAODING: A High-fidelity Data-free Pruning Toolkit for Debloating Pre-trained Neural Networks

We present PAODING, a toolkit to debloat pretrained neural network models through the lens of data-free pruning. To preserve the model fidelity, PAODING adopts an iterative process, which dynamically measures the effect of deleting a neuron to identify candidates that have the least impact to the output layer. Our evaluation shows that PAODING can significantly reduce the model size, generalize on different datasets and models, and meanwhile preserve the model fidelity in terms of test accuracy and adversarial robustness. PAODING is publicly available on PyPI via https://pypi.org/project/paoding-dl.

cs.LG↗

arXiv2025

Assessing Privacy Compliance of Android Third-Party SDKs

Third-party Software Development Kits (SDKs) are widely adopted in Android app development, to effortlessly accelerate development pipelines and enhance app functionality. However, this convenience raises substantial concerns about unauthorized access to users' privacy-sensitive information, which could be further abused for illegitimate purposes like user tracking or monetization. Our study offers a targeted analysis of user privacy protection among Android third-party SDKs, filling a critical gap in the Android software supply chain. It focuses on two aspects of their privacy practices, including data exfiltration and behavior-policy compliance (or privacy compliance), utilizing techniques of taint analysis and large language models. It covers 158 widely-used SDKs from two key SDK release platforms, the official one and a large alternative one. From them, we identified 338 instances of privacy data exfiltration. On the privacy compliance, our study reveals that more than 30% of the examined SDKs fail to provide a privacy policy to disclose their data handling practices. Among those that provide privacy policies, 37% of them over-collect user data, and 88% falsely claim access to sensitive data. We revisit the latest versions of the SDKs after 12 months. Our analysis demonstrates a persistent lack of improvement in these concerning trends. Based on our findings, we propose three actionable recommendations to mitigate the privacy leakage risks and enhance privacy protection for Android users. Our research not only serves as an urgent call for industry attention but also provides crucial insights for future regulatory interventions.

cs.CR↗