Searcharxiv⌕ Search

arXiv subjects

Xiaoyong Yuan

Publications and source records attributed to Xiaoyong Yuan.

At least 19 recordsLinked to original sources
arXiv2026

No Concept Escapes the Audit: Auditing-Aware Unlearning for Verifiable Concept Erasure in Diffusion Models

Text-to-image diffusion models can generate prohibited content, which motivates concept erasure through machine unlearning. Most erasure methods intervene at the text interface, through prompt modification or localized updates to text-conditioning weights, and they are evaluated by what the model outputs for given prompts. Such evaluation cannot see what the network still encodes. Latent-space auditing, which bypasses text conditioning and probes the denoising network directly, shows that erased concepts remain recoverable from internal representations. We find that this also holds for methods built to be robust against adversarial prompts, and that the problem grows with the number of erased concepts. We propose Auditing-Aware Unlearning for Verifiable Concept Erasure in Diffusion Models (AVCE), a framework that grounds erasure in the model's latent representations. AVCE audits the embedding neighborhood of each concept and condenses the discovered vulnerable directions into an anchor at the weakest geometric point. It edits cross-attention and self-attention projections in closed form at this anchor, then fine-tunes the two pathways with pathway-level auditing losses, using orthogonal gradient projection to consolidate multiple concepts. Experiments on SD v1.5, SDXL, and Flux 1.0 across object, explicit-content, and artistic-style unlearning show that AVCE reduces attack success rates by 5.07x and improves auditing scores by 3.84x over the strongest baseline, while preserving competitive generation quality.

cs.LG↗
arXiv2026

DiffRadar: Differentiable Physics-Aware Radar SLAM with Gaussian Fields

Radar sensing is increasingly used in mobile systems because it operates reliably under poor lighting, adverse weather, and privacy-sensitive settings where cameras and LiDAR often fail. However, most existing radar SLAM systems estimate motion through scan matching on discretized radar heatmaps, which breaks geometric continuity and fails to capture key radar sensing properties, often leading to unstable pose estimation and degraded mapping in regenerate or dynamically changing environments. We present DiffRadar, a real-time radar SLAM system that models radar observations as a differentiable, physics-aware Gaussian field rather than discrete scans. DiffRadar represents the scene as anisotropic Gaussian primitives and renders radar measurements in range-azimuth and Doppler-azimuth spaces through a differentiable radar forward model, enabling joint optimization of robot pose and scene structure directly from radar measurements. We implement DiffRadar on commodity FMCW radar hardware and evaluate it on both the public Radarize benchmark and a controlled stress-test suite that targets common radar SLAM failure modes, including corridor degeneracy, motion regime transitions, dynamic clutter, and long-horizon loop closures. DiffRadar achieves substantial reductions in trajectory error on the benchmark, with especially large gains under feature-poor corridor motion, while more than doubling map consistency and maintaining real-time performance at 70 FPS. These results show that modeling radar observations directly in the signal domain enables substantially more robust and consistent radar-only SLAM for mobile platforms.

cs.RO↗
arXiv2026

When AI Persuades: Adversarial Explanation Attacks on Human Trust in AI-Assisted Decision Making

Most adversarial threats in artificial intelligence (AI) target the computational behavior of models rather than the humans who rely on them. Yet modern AI systems increasingly operate within human decision loops, where users interpret and act on model recommendations. Large Language Models (LLMs) generate fluent natural-language explanations that shape how users perceive and trust AI outputs, revealing a new attack surface at the cognitive layer: the communication channel between AI and its users. We introduce adversarial explanation attacks (AEAs), where an attacker manipulates the framing of LLM-generated explanations to modulate human trust in incorrect outputs. We formalize this behavioral threat through the trust miscalibration gap, a metric that captures the difference in human trust between benign and adversarial explanations. Using this metric as a lens, we highlight a behavioral risk where persuasive explanation framing can preserve user trust even when the underlying AI prediction is wrong. To characterize this threat, we conducted a human study with over 200 participants, systematically varying four dimensions of explanation framing: reasoning mode, evidence type, communication style, and presentation format. Our findings show that users report nearly identical trust for adversarial and benign explanations, with adversarial explanations preserving the vast majority of benign trust despite being incorrect. The most vulnerable cases arise when AEAs closely resemble expert communication, combining authoritative evidence, neutral tone, and domain-appropriate reasoning. Vulnerability is highest on hard tasks, in fact-driven domains, and among participants who are less formally educated, younger, or highly trusting of AI.

cs.AI↗
arXiv2026

OracleTSC: Oracle-Informed Reward Hurdle and Uncertainty Regularization for Traffic Signal Control

Transparent decision-making is essential for traffic signal control (TSC) systems to earn public trust. However, traditional reinforcement learning-based TSC methods function as black boxes with limited interpretability. Although large language models (LLMs) can provide natural language reasoning, reinforcement finetuning for TSC remains unstable because feedback is sparse and delayed, while most actions produce only marginal changes in congestion metrics. We introduce OracleTSC, which stabilizes LLM-based TSC through two mechanisms: (1) a reward hurdle mechanism that filters weak learning signals by subtracting a calibrated threshold from environmental rewards, and (2) uncertainty regularization that maximizes the probability of the selected response to encourage consistent decisions across sampled outputs. Experiments on the LibSignal benchmark show that OracleTSC enables a compact LLaMA3-8B model to substantially improve traffic efficiency, achieving a 75% reduction in travel time and a 67% decrease in queue length compared with the pretrained baseline while preserving interpretability through natural language explanations. OracleTSC also demonstrates strong cross-intersection generalization: a policy trained on one intersection transfers to a structurally different intersection with 17% lower travel time and 39% lower queue length without additional finetuning. These results suggest that uncertainty-aware reward shaping can improve the stability and effectiveness of reinforcement fine-tuning for TSC.

cs.AI↗
arXiv2026

Secure Semantic Communications via AI Defenses: Fundamentals, Solutions, and Future Directions

Semantic communication (SemCom) redefines wireless communication from reproducing symbols to transmitting task-relevant semantics. However, this AI-native architecture also introduces new vulnerabilities, as semantic failures may arise from adversarial perturbations to models, corrupted training data, desynchronized priors, or misaligned inference even when lower-layer transmission reliability and cryptographic protection remain intact. This survey provides a defense-centered and system-oriented synthesis of security in SemCom via AI defense. We analyze AI-centric threat models by consolidating existing studies and organizing attack surfaces across model-level, channel-realizable, knowledge-based, and networked inference vectors. Building on this foundation, we present a structured taxonomy of defense strategies organized by where semantic integrity can be compromised in SemCom systems despite correct symbol delivery, spanning semantic encoding, wireless transmission, knowledge integrity, and coordination among multiple agents. These categories correspond to distinct security failure modes, including representation fragility, channel-realizable manipulation, semantic prior poisoning or desynchronization, and adversarial propagation through distributed inference. We also examine security utility operating envelopes that capture tradeoffs among semantic fidelity, robustness, latency, and energy under realistic constraints, survey evaluation frameworks and representative applications, and identify open challenges in cross-layer composition and deployment-time certification. Overall, this survey offers a unified system-level perspective that enables readers to understand major threat and defense mechanisms in AI-native SemCom systems and to leverage emerging security techniques in the design and deployment of robust SemCom architectures for next-generation intelligent networks.

cs.CR↗
arXiv2025

A Simple Framework Towards Vision-based Traffic Signal Control with Microscopic Simulation

Traffic signal control (TSC) is crucial for reducing traffic congestion leading to smoother traffic flow, reduced idle time, and mitigated CO2 emissions. In this paper, we explore the computer vision approach for TSC that modulates on-road traffic flows through visual observation. Unlike traditional feature-based approaches, vision-based methods depend much less on heuristics and predefined features, bringing promising potentials for end-to-end learning and optimization of traffic signals. Thus, we introduce a simple traffic simulation framework called TrafficDojo towards vision-based TSC and its benchmark by integrating the microscopic traffic flow provided in SUMO into the 3D driving simulator MetaDrive. This proposed framework offers a versatile traffic environment for in-depth analysis and comprehensive evaluation of traffic signal controllers across diverse traffic conditions and scenarios. We establish and compare baseline algorithms including both traditional and Reinforcement Learning (RL) approaches. This work sheds light on the design and development of vision-based TSC approaches and opens up new research opportunities

cs.CV↗
arXiv2025

MOBA: A Material-Oriented Backdoor Attack against LiDAR-based 3D Object Detection Systems

LiDAR-based 3D object detection is widely used in safety-critical systems. However, these systems remain vulnerable to backdoor attacks that embed hidden malicious behaviors during training. A key limitation of existing backdoor attacks is their lack of physical realizability, primarily due to the digital-to-physical domain gap. Digital triggers often fail in real-world settings because they overlook material-dependent LiDAR reflection properties. On the other hand, physically constructed triggers are often unoptimized, leading to low effectiveness or easy detectability.This paper introduces Material-Oriented Backdoor Attack (MOBA), a novel framework that bridges the digital-physical gap by explicitly modeling the material properties of real-world triggers. MOBA tackles two key challenges in physical backdoor design: 1) robustness of the trigger material under diverse environmental conditions, 2) alignment between the physical trigger's behavior and its digital simulation. First, we propose a systematic approach to selecting robust trigger materials, identifying titanium dioxide (TiO_2) for its high diffuse reflectivity and environmental resilience. Second, to ensure the digital trigger accurately mimics the physical behavior of the material-based trigger, we develop a novel simulation pipeline that features: (1) an angle-independent approximation of the Oren-Nayar BRDF model to generate realistic LiDAR intensities, and (2) a distance-aware scaling mechanism to maintain spatial consistency across varying depths. We conduct extensive experiments on state-of-the-art LiDAR-based and Camera-LiDAR fusion models, showing that MOBA achieves a 93.50% attack success rate, outperforming prior methods by over 41%. Our work reveals a new class of physically realizable threats and underscores the urgent need for defenses that account for material-level properties in real-world environments.

cs.CV↗
arXiv2025

What Lurks Within? Concept Auditing for Shared Diffusion Models at Scale

Diffusion models (DMs) have revolutionized text-to-image generation, enabling the creation of highly realistic and customized images from text prompts. With the rise of parameter-efficient fine-tuning (PEFT) techniques, users can now customize powerful pre-trained models using minimal computational resources. However, the widespread sharing of fine-tuned DMs on open platforms raises growing ethical and legal concerns, as these models may inadvertently or deliberately generate sensitive or unauthorized content. Despite increasing regulatory attention on generative AI, there are currently no practical tools for systematically auditing these models before deployment. In this paper, we address the problem of concept auditing: determining whether a fine-tuned DM has learned to generate a specific target concept. Existing approaches typically rely on prompt-based input crafting and output-based image classification but they suffer from critical limitations, including prompt uncertainty, concept drift, and poor scalability. To overcome these challenges, we introduce Prompt-Agnostic Image-Free Auditing (PAIA), a novel, model-centric concept auditing framework. By treating the DM as the object of inspection, PAIA enables direct analysis of internal model behavior, bypassing the need for optimized prompts or generated images. We evaluate PAIA on 320 controlled models trained with curated concept datasets and 771 real-world community models sourced from a public DM sharing platform. Evaluation results show that PAIA achieves over 90% detection accuracy while reducing auditing time by 18 - 40X compared to existing baselines. To our knowledge, PAIA is the first scalable and practical solution for pre-deployment concept auditing of diffusion models, providing a practical foundation for safer and more transparent diffusion model sharing.

cs.LG↗
arXiv2025

Who's Wearing? Ear Canal Biometric Key Extraction for User Authentication on Wireless Earbuds

Ear canal scanning/sensing (ECS) has emerged as a novel biometric authentication method for mobile devices paired with wireless earbuds. Existing studies have demonstrated the uniqueness of ear canals by training and testing machine learning classifiers on ECS data. However, implementing practical ECS-based authentication requires preventing raw biometric data leakage and designing computationally efficient protocols suitable for resource-constrained earbuds. To address these challenges, we propose an ear canal key extraction protocol, \textbf{EarID}. Without relying on classifiers, EarID extracts unique binary keys directly on the earbuds during authentication. These keys further allow the use of privacy-preserving fuzzy commitment scheme that verifies the wearer's key on mobile devices. Our evaluation results demonstrate that EarID achieves a 98.7\% authentication accuracy, comparable to machine learning classifiers. The mobile enrollment time (160~ms) and earbuds processing time (226~ms) are negligible in terms of wearer's experience. Moreover, our approach is robust and attack-resistant, maintaining a false acceptance rate below 1\% across all adversarial scenarios. We believe the proposed EarID offers a practical and secure solution for next-generation wireless earbuds.

cs.CR↗
arXiv2025

Your RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor Attacks

Retrieval-augmented generation (RAG) enhances factual grounding by integrating retrieval mechanisms with generative models but introduces new attack surfaces, particularly through backdoor attacks. While prior research has largely focused on disinformation threats, fairness vulnerabilities remain underexplored. Unlike conventional backdoors that rely on direct trigger-to-target mappings, fairness-driven attacks exploit the interaction between retrieval and generation models, manipulating semantic relationships between target groups and social biases to establish a persistent and covert influence on content generation. This paper introduces BiasRAG, a systematic framework that exposes fairness vulnerabilities in RAG through a two-phase backdoor attack. During the pre-training phase, the query encoder is compromised to align the target group with the intended social bias, ensuring long-term persistence. In the post-deployment phase, adversarial documents are injected into knowledge bases to reinforce the backdoor, subtly influencing retrieved content while remaining undetectable under standard fairness evaluations. Together, BiasRAG ensures precise target alignment over sensitive attributes, stealthy execution, and resilience. Empirical evaluations demonstrate that BiasRAG achieves high attack success rates while preserving contextual relevance and utility, establishing a persistent and evolving threat to fairness in RAG.

cs.IR↗
arXiv2025

DyME: Dynamic Multi-Concept Erasure in Diffusion Models with Bi-Level Orthogonal LoRA Adaptation

Text-to-image diffusion models (DMs) inadvertently reproduce copyrighted styles and protected visual concepts, raising legal and ethical concerns. Concept erasure has emerged as a safeguard, aiming to selectively suppress such concepts through fine-tuning. However, existing methods do not scale to practical settings where providers must erase multiple and possibly conflicting concepts. The core bottleneck is their reliance on static erasure: a single checkpoint is fine-tuned to remove all target concepts, regardless of the actual erasure needs at inference. This rigid design mismatches real-world usage, where requests vary per generation, leading to degraded erasure success and reduced fidelity for non-target content. We propose DyME, an on-demand erasure framework that trains lightweight, concept-specific LoRA adapters and dynamically composes only those needed at inference. This modular design enables flexible multi-concept erasure, but naive composition causes interference among adapters, especially when many or semantically related concepts are suppressed. To overcome this, we introduce bi-level orthogonality constraints at both the feature and parameter levels, disentangling representation shifts and enforcing orthogonal adapter subspaces. We further develop ErasureBench-H, a new hierarchical benchmark with brand-series-character structure, enabling principled evaluation across semantic granularities and erasure set sizes. Experiments on ErasureBench-H and standard datasets (e.g., CIFAR-100, Imagenette) demonstrate that DyME consistently outperforms state-of-the-art baselines, achieving higher multi-concept erasure fidelity with minimal collateral degradation.

cs.CV↗
arXiv2025

Position: Human Factors Reshape Adversarial Analysis in Human-AI Decision-Making Systems

As Artificial Intelligence (AI) increasingly supports human decision-making, its vulnerability to adversarial attacks grows. However, the existing adversarial analysis predominantly focuses on fully autonomous AI systems, where decisions are executed without human intervention. This narrow focus overlooks the complexities of human-AI collaboration, where humans interpret, adjust, and act upon AI-generated decisions. Trust, expectations, and cognitive behaviors influence how humans interact with AI, creating dynamic feedback loops that adversaries can exploit. To strengthen the robustness of AI-assisted decision-making, adversarial analysis must account for the interplay between human factors and attack strategies. This position paper argues that human factors fundamentally reshape adversarial analysis and must be incorporated into evaluating robustness in human-AI decision-making systems. To fully explore human factors in adversarial analysis, we begin by investigating the role of human factors in human-AI collaboration through a comprehensive review. We then introduce a novel robustness analysis framework that (1) examines how human factors affect collaborative decision-making performance, (2) revisits and interprets existing adversarial attack strategies in the context of human-AI interaction, and (3) introduces a new timing-based adversarial attack as a case study, illustrating vulnerabilities emerging from sequential human actions. The experimental results reveal that attack timing uniquely impacts decision outcomes in human-AI collaboration. We hope this analysis inspires future research on adversarial robustness in human-AI systems, fostering interdisciplinary approaches that integrate AI security, human cognition, and decision-making dynamics.

cs.HC↗
arXiv2025

AIP: Subverting Retrieval-Augmented Generation via Adversarial Instructional Prompt

Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by retrieving relevant documents from external sources to improve factual accuracy and verifiability. However, this reliance introduces new attack surfaces within the retrieval pipeline, beyond the LLM itself. While prior RAG attacks have exposed such vulnerabilities, they largely rely on manipulating user queries, which is often infeasible in practice due to fixed or protected user inputs. This narrow focus overlooks a more realistic and stealthy vector: instructional prompts, which are widely reused, publicly shared, and rarely audited. Their implicit trust makes them a compelling target for adversaries to manipulate RAG behavior covertly. We introduce a novel attack for Adversarial Instructional Prompt (AIP) that exploits adversarial instructional prompts to manipulate RAG outputs by subtly altering retrieval behavior. By shifting the attack surface to the instructional prompts, AIP reveals how trusted yet seemingly benign interface components can be weaponized to degrade system integrity. The attack is crafted to achieve three goals: (1) naturalness, to evade user detection; (2) utility, to encourage use of prompts; and (3) robustness, to remain effective across diverse query variations. We propose a diverse query generation strategy that simulates realistic linguistic variation in user queries, enabling the discovery of prompts that generalize across paraphrases and rephrasings. Building on this, a genetic algorithm-based joint optimization is developed to evolve adversarial prompts by balancing attack success, clean-task utility, and stealthiness. Experimental results show that AIP achieves up to 95.23% ASR while preserving benign functionality. These findings uncover a critical and previously overlooked vulnerability in RAG systems, emphasizing the need to reassess the shared instructional prompts.

cs.CV↗
arXiv2025

Sample-Efficient Reinforcement Learning Controller for Deep Brain Stimulation in Parkinson's Disease

Deep brain stimulation (DBS) is an established intervention for Parkinson's disease (PD), but conventional open-loop systems lack adaptability, are energy-inefficient due to continuous stimulation, and provide limited personalization to individual neural dynamics. Adaptive DBS (aDBS) offers a closed-loop alternative, using biomarkers such as beta-band oscillations to dynamically modulate stimulation. While reinforcement learning (RL) holds promise for personalized aDBS control, existing methods suffer from high sample complexity, unstable exploration in binary action spaces, and limited deployability on resource-constrained hardware. We propose SEA-DBS, a sample-efficient actor-critic framework that addresses the core challenges of RL-based adaptive neurostimulation. SEA-DBS integrates a predictive reward model to reduce reliance on real-time feedback and employs Gumbel Softmax-based exploration for stable, differentiable policy updates in binary action spaces. Together, these components improve sample efficiency, exploration robustness, and compatibility with resource-constrained neuromodulatory hardware. We evaluate SEA-DBS on a biologically realistic simulation of Parkinsonian basal ganglia activity, demonstrating faster convergence, stronger suppression of pathological beta-band power, and resilience to post-training FP16 quantization. Our results show that SEA-DBS offers a practical and effective RL-based aDBS framework for real-time, resource-constrained neuromodulation.

cs.LG↗
arXiv2025

You Don't Need All Attentions: Distributed Dynamic Fine-Tuning for Foundation Models

Fine-tuning plays a crucial role in adapting models to downstream tasks with minimal training efforts. However, the rapidly increasing size of foundation models poses a daunting challenge for accommodating foundation model fine-tuning in most commercial devices, which often have limited memory bandwidth. Techniques like model sharding and tensor parallelism address this issue by distributing computation across multiple devices to meet memory requirements. Nevertheless, these methods do not fully leverage their foundation nature in facilitating the fine-tuning process, resulting in high computational costs and imbalanced workloads. We introduce a novel Distributed Dynamic Fine-Tuning (D2FT) framework that strategically orchestrates operations across attention modules based on our observation that not all attention modules are necessary for forward and backward propagation in fine-tuning foundation models. Through three innovative selection strategies, D2FT significantly reduces the computational workload required for fine-tuning foundation models. Furthermore, D2FT addresses workload imbalances in distributed computing environments by optimizing these selection strategies via multiple knapsack optimization. Our experimental results demonstrate that the proposed D2FT framework reduces the training computational costs by 40% and training communication costs by 50% with only 1% to 2% accuracy drops on the CIFAR-10, CIFAR-100, and Stanford Cars datasets. Moreover, the results show that D2FT can be effectively extended to recent LoRA, a state-of-the-art parameter-efficient fine-tuning technique. By reducing 40% computational cost or 50% communication cost, D2FT LoRA top-1 accuracy only drops 4% to 6% on Stanford Cars dataset.

cs.LG↗
arXiv2024

BadFusion: 2D-Oriented Backdoor Attacks against 3D Object Detection

3D object detection plays an important role in autonomous driving; however, its vulnerability to backdoor attacks has become evident. By injecting ''triggers'' to poison the training dataset, backdoor attacks manipulate the detector's prediction for inputs containing these triggers. Existing backdoor attacks against 3D object detection primarily poison 3D LiDAR signals, where large-sized 3D triggers are injected to ensure their visibility within the sparse 3D space, rendering them easy to detect and impractical in real-world scenarios. In this paper, we delve into the robustness of 3D object detection, exploring a new backdoor attack surface through 2D cameras. Given the prevalent adoption of camera and LiDAR signal fusion for high-fidelity 3D perception, we investigate the latent potential of camera signals to disrupt the process. Although the dense nature of camera signals enables the use of nearly imperceptible small-sized triggers to mislead 2D object detection, realizing 2D-oriented backdoor attacks against 3D object detection is non-trivial. The primary challenge emerges from the fusion process that transforms camera signals into a 3D space, compromising the association with the 2D trigger to the target output. To tackle this issue, we propose an innovative 2D-oriented backdoor attack against LiDAR-camera fusion methods for 3D object detection, named BadFusion, for preserving trigger effectiveness throughout the entire fusion process. The evaluation demonstrates the effectiveness of BadFusion, achieving a significantly higher attack success rate compared to existing 2D-oriented attacks.

cs.CV↗
arXiv2024

Improving Channel Resilience for Task-Oriented Semantic Communications: A Unified Information Bottleneck Approach

Task-oriented semantic communications (TSC) enhance radio resource efficiency by transmitting task-relevant semantic information. However, current research often overlooks the inherent semantic distinctions among encoded features. Due to unavoidable channel variations from time and frequency-selective fading, semantically sensitive feature units could be more susceptible to erroneous inference if corrupted by dynamic channels. Therefore, this letter introduces a unified channel-resilient TSC framework via information bottleneck. This framework complements existing TSC approaches by controlling information flow to capture fine-grained feature-level semantic robustness. Experiments on a case study for real-time subchannel allocation validate the framework's effectiveness.

cs.IT↗
arXiv2023

PATROL: Privacy-Oriented Pruning for Collaborative Inference Against Model Inversion Attacks

Collaborative inference has been a promising solution to enable resource-constrained edge devices to perform inference using state-of-the-art deep neural networks (DNNs). In collaborative inference, the edge device first feeds the input to a partial DNN locally and then uploads the intermediate result to the cloud to complete the inference. However, recent research indicates model inversion attacks (MIAs) can reconstruct input data from intermediate results, posing serious privacy concerns for collaborative inference. Existing perturbation and cryptography techniques are inefficient and unreliable in defending against MIAs while performing accurate inference. This paper provides a viable solution, named PATROL, which develops privacy-oriented pruning to balance privacy, efficiency, and utility of collaborative inference. PATROL takes advantage of the fact that later layers in a DNN can extract more task-specific features. Given limited local resources for collaborative inference, PATROL intends to deploy more layers at the edge based on pruning techniques to enforce task-specific features for inference and reduce task-irrelevant but sensitive features for privacy preservation. To achieve privacy-oriented pruning, PATROL introduces two key components: Lipschitz regularization and adversarial reconstruction training, which increase the reconstruction errors by reducing the stability of MIAs and enhance the target inference model by adversarial training, respectively. On a real-world collaborative inference task, vehicle re-identification, we demonstrate the superior performance of PATROL in terms of against MIAs.

cs.LG↗