Searcharxiv⌕ Search

arXiv subjects

Qingzhao Zhang

Publications and source records attributed to Qingzhao Zhang.

At least 19 recordsLinked to original sources
arXiv2026

Investigating Adversarial Robustness of Heterogeneous Cooperative Perception

Heterogeneous cooperative perception (CP) enables connected vehicles with diverse sensor setups to share spatial awareness via compact feature maps, where receivers reconcile these maps using learned translation modules for fusion and inference. Prior attacks against CP in a homogeneous setting reveal that the data exchange introduces a critical attack surface: a single malicious agent can transmit crafted features that erase real objects from a neighbor's fused scene. Yet, it is widely hypothesized that heterogeneity naturally defends against these attacks, as the attacker lacks knowledge of the victim's detector and the translation module scrambles adversarial gradients. We demonstrate that this protection is largely an illusion. Using a matched-objective harness to standardize the perturbation budget, objective, and forward path, we show that properly tuned iterative attacks close or reverse the apparent robustness gap. However, these optimization-based attacks require ground-truth labels and iterative backpropagation, meaning they do not represent a practical field threat running in real-time. To bridge this gap, we introduce HetPoison, a learned generator that crafts a removal perturbation in a single, label-free forward pass. HetPoison transfers across major heterogeneous designs without requiring access to the victim's detector, matching or exceeding the effectiveness of expensive optimizer-based attacks. Since heterogeneity itself is not a defense, we propose HetShield, a lightweight trust layer that validates the spatiotemporal consistency across features, recovering 83--95% of the accuracy degraded by attacks, outperforming prior art.

cs.CV↗
arXiv2026

From Stealthy Data Fabrication to Unsafe Driving: Realistic Scenario Attacks on Collaborative Perception

Collaborative perception allows connected and autonomous vehicles (CAVs) to improve perception by sharing sensory data, but it also introduces security risks from manipulated inputs. Prior work shows that attackers can spoof or remove objects by fabricating shared data, yet the practicality of such attacks in real-world driving remains unclear. Existing attacks are often detectable or evaluated in manually constructed scenarios, leaving open whether they can induce safety-critical outcomes in dynamic environments. To bridge this gap, we present a stealthy, scenario-realistic data fabrication attack that induces unsafe driving behaviors through end-to-end system effects. Instead of creating large, easily detectable anomalies, our attack subtly manipulates the poses of existing objects in shared perception results, keeping perturbations below detection thresholds. These small errors are then propagated through downstream modules, including object tracking and trajectory prediction, leading to significant deviations in predicted behaviors and ultimately unsafe driving decisions. We further design an online, scenario-aware attack framework that adapts to dynamic traffic conditions and optimizes attack strategies at runtime. Experiments on OPV2V and V2X-Real demonstrate that the attack achieves over 90% success in inducing detection errors and triggers safety-critical behaviors, such as unnecessary hard braking, in up to 50% of scenarios, while largely evading state-of-the-art defenses. We also propose a mitigation that focuses on detecting anomalies in localized, safety-critical regions, achieving an 80% detection rate on the small pose perturbation compared to 11% for the best existing methods.

cs.CR↗
arXiv2026

SoK: How Sensor Attacks Disrupt Autonomous Vehicles: An End-to-end Analysis, Challenges, and Missed Threats

Autonomous vehicles, including self-driving cars, ground robots, and drones, rely on multi-modal sensor pipelines for safe operation, yet remain vulnerable to adversarial sensor attacks. A critical gap is the lack of a systematic end-to-end view of how sensor-induced errors traverse interconnected modules to affect the physical world. To address this gap, we provide a comprehensive survey across platforms, sensing modalities, attack methods, and countermeasures. At its core is Systematic Error Propagation Graph (SEPG), a graph-based framework that maps how attacks inject errors, the conditions for their propagation through modules from perception and localization to planning and control, and when they reach physical impact. From this analysis, our study distills 8 key findings that highlight the feasibility constraints of sensor attacks and uncovers 12 previously overlooked attack vectors exploiting inter-module interactions, several of which we validate via proof-of-concept experiments. We further conduct a pilot user study with autonomous vehicle security researchers, showing that SEPG is viewed as a faithful and useful framework for qualitative system-level attack analysis. We also evaluate large language models (LLMs) for generating the analysis, demonstrating the potential of AI-powered automation.

cs.CR↗
arXiv2026

Adversarial Trust Poisoning in Vehicular Collaborative Perception

Collaborative perception (CP) enables connected and autonomous vehicles to share sensor data and jointly reason about their environment. To defend against adversaries that fabricate or manipulate shared data, existing systems employ cross-vehicle inconsistency detection and trust estimation, penalizing vehicles whose observations conflict with the majority. In this work, we show that these defenses themselves introduce a new attack surface. We present TrustFlip, a novel attack that weaponizes consistency-based defenses to poison the trust assigned to benign vehicles. Instead of injecting false data into the collaboration pipeline, it deploys physical adversarial objects that are genuine but induce inconsistent observations among benign vehicles. The resulting inconsistencies are misattributed by the defense to the targeted vehicle, causing its trust score to degrade and eventually leading to its downweighting or exclusion from collaboration. Consequently, the system loses reliable sensing contributors, degrading perception capability and potentially inducing safety-critical failures. We evaluate TrustFlip across multiple collaborative perception architectures and defense mechanisms. Our results show that state-of-the-art defenses can be significantly affected: the attack removes the targeted benign vehicle from collaboration in up to 87.7% of scenarios and drops Average Precision (AP) by up to 13%. As an initial mitigation, we introduce TrustReflect, a lightweight self-reflection mechanism that marks disputed regions as uncertain and excludes them from trust evaluation, reducing the attack success rate by 35-100%.

cs.CR↗
arXiv2026

Adversarial Calibration Attack on Autonomous Vehicles

Autonomous vehicles (AVs) rely on accurate camera-LiDAR calibration for multimodal sensor fusion. In practice, calibration can drift due to vibration, temperature variation, or minor sensor displacement, motivating online calibration algorithms that detect and correct misalignment at runtime while allowing the vehicle to continue operating without a factory visit. Existing AV attacks largely assume correct calibration. We instead identify online sensor calibration as a new attack plane. A corrupted calibration update can persist across subsequent fusion operations, causing system-wide errors that propagate from perception to planning and control. We present Adversarial Calibration Attack (ACA), the first physical attack against camera-LiDAR online calibration. Using a single adversarial poster, ACA first spoofs the miscalibration detector to trigger the calibration process and then steers the calibration estimator toward an incorrect transformation. A unified optimization jointly designs the poster's geometry and texture for both objectives. We evaluate ACA across benchmark datasets, simulation, and physical experiments. On benchmark datasets such as KITTI and nuScenes, ACA induces up to 33.9 degrees mean rotational calibration error, thereby severely degrading object detection. In the CARLA simulator, the attack causes a collision when the corrupted calibration is accepted in vulnerable scenarios crafted by the attacker. On a real Husky robot, a printed adversarial poster successfully reproduces the calibration error. These results demonstrate that online calibration is a practical and safety-critical attack surface for AVs.

cs.RO↗
arXiv2026

Local spectral clustering for heterogeneous clustering structures

Classical clustering methods typically assume that all informative features support a single latent partition of the observations. This assumption can be overly restrictive for modern high-dimensional data, where different subsets of features may encode distinct notions of similarity and induce heterogeneous sample partitions, while some features may contain no meaningful clustering information. We develop a frequentist framework for local clustering that simultaneously identifies feature groups and estimates the sample clustering structure associated with each group. Our approach represents each sample partition by a label-invariant clustering matrix and groups features according to their shared clustering structures, thereby reformulating local clustering as a feature-grouping, or clustering-of-clusterings, problem. Under a heterogeneous sub-Gaussian mixture model, we construct feature-specific Gaussian-kernel similarity matrices and propose a local spectral clustering procedure based on a clustering-matrix optimization criterion. The proposed method avoids explicit likelihood specification and Bayesian posterior computation, accommodates heterogeneous feature distributions, and permits the presence of non-informative features. Extensive simulations and applications further demonstrate the practical utility and superiority of the proposed approach.

stat.ME↗
arXiv2026

Revisiting the Adversarial Robustness of Graph-Based Traffic Forecasting

Traffic forecasting by graph-based AI is a critical component of intelligent transportation systems, motivating security research on robustness to malicious sensor readings. We argue that prior robustness evaluations are largely shaped by unrealistic threat models and untargeted objectives, so both attacks and defenses must be revisited. We study a practical adversary with limited model knowledge and the ability to monitor and manipulate only a few road sensors. More importantly, practical attacks can be localized to specific links or routes, causing incorrect estimated arrival times or unnecessary rerouting while leaving the broader network largely unaffected. This targeted setting remains underexplored, and defenses such as adversarial training do not transfer well from the norm-bounded attacks they train on to structurally different, physics-aware attacks that mimic genuine congestion. We therefore reframe robustness as a detection problem, introducing a learned physics-informed detector whose output is fed to a hardened forecaster as an input feature and trained against adaptive attacks with the forecaster fixed. We evaluate across a variety of model architectures and benchmarks. The physics-aware attack multiplies target-link error several-fold while the network-wide error barely moves, and adversarial training, tuned to norm-bounded perturbations, barely dents it. Our detection--mitigation defense improves even on adversarial training hardened against the physics-aware attack itself, on $13$ of $15$ model--dataset settings and by the widest margin on a held-out attack, at near-zero clean cost. The results emphasize the need to examine abstracted AI adversarial attacks under application-specific constraints to assess their true security impacts.

cs.CR↗
arXiv2026

Banshee: Target Switch Attacks on Gimbal-Stabilized Visual Tracking Systems via Acoustic Injection

Gimbal-stabilized visual tracking is critical for modern autonomous systems such as Unmanned Aerial Vehicles (UAVs). While prior work shows acoustic signals can disturb gimbal internals, the impact of such attacks on real-world applications like UAV tracking and following remains underexplored. Existing demonstrations largely overlook practical challenges for real-world attacks, such as object-motion uncertainty and runtime latency. To bridge this gap, we present Banshee, the first physically realizable attack that induces target switching in UAV visual tracking systems by exploiting acoustic vulnerabilities in gimbal-camera systems. Banshee generates carefully crafted acoustic waveforms that induce optimized adversarial gimbal oscillations, causing directionally biased camera-view drifts that break inter-frame target associations. Consequently, the onboard tracker is driven to switch from the original target to an attacker-selected object with high probability, with occasional target loss. Banshee achieves a 93.6% success rate in simulation across two commercial gimbal systems and five trackers. Real-world benchtop and in-flight black-box attacks against a commercial drone across varied scenarios show an overall 95.5% attack success rate. Our results reveal a practical cross-domain vulnerability between acoustics and vision, highlighting the need for robust designs of gimbal systems and applications. Our code is available at: https://github.com/U1ltra/Banshee.

cs.CV↗
arXiv2026

CLAP: Contrastive Latent-space Prompt Optimization for End-to-end Autonomous Driving

End-to-end autonomous driving systems powered by Vision-Language-Action (VLA) models achieve strong performance on common driving scenarios, yet remain brittle in rare but safety-critical long-tail situations such as active construction zones and complex yielding geometries. In this paper, we present a method that addresses the long-tail challenging scenes beyond data scaling and model training. We introduce CLAP (Contrastive Latent-space Prompt optimization), a location-aware adaptation framework that augments a frozen VLA driving model with per-roadblock soft prompts, optimized from crowdsourced data and retrieved on demand via Vehicle-to-Everything (V2X) communication. Our approach rests on two observations from VLAs' latent space: (i) at the VLA's hidden-state layer, scenarios from the same roadblock cluster tightly and occupy compact regions of the latent space; and (ii) within a single roadblock, long-tail and normal frames are heavily intermixed in the latent representation, making it difficult to improve one without disturbing the other. CLAP addresses this via a two-stage pipeline: supervised contrastive learning to discover a roadblock-specific hard-scene direction, followed by directionally regularized prompt optimization that selectively improves challenging frames while preserving normal frame performance. On the NAVSIM benchmark with various state-of-the-art VLA backbones, CLAP reduces challenging scenario planning error by 24% with no regression on normal frames, significantly improving planning performance.

cs.CV↗
arXiv2026

Still Camouflage, Moving Illusion: View-Induced Trajectory Manipulation in Autonomous Driving

Existing physical adversarial attacks on vision-based autonomous driving induce time-evolving perception errors, including biased object tracking or trajectory prediction, through (i) sophisticated physical patch inducing detection box drift when entering the view distance, or (ii) dynamically changing patches that cause different perception errors at different time. In both cases, viewing-angle variation is treated as a challenge, requiring adversarial patches to remain effective across frames under varying views, leading to complex multi-view optimization. In contrast, we show that viewing-angle variation itself can be turned into an attack tool. We design a new attack paradigm where a static, passive adversarial camouflage is mounted on a vehicle whose view-dependent appearance naturally evolves with relative motion, inducing consistent feature drift across frames. This causes the system to infer a physically plausible but incorrect trajectory, such as a false cut-in, which propagates to downstream decision-making and triggers unnecessary braking. Unlike prior approaches that require multi-view robustness or active intervention, our attack emerges from normal driving dynamics and is easy to deploy: a parked vehicle with a natural camouflage can induce hard braking in passing autonomous vehicles. We demonstrate the novel attack on nuScenes dataset, showing the effectiveness with an end-to-end success rate of up to 87.5%, measured by hard-braking events, and robustness across different scene backgrounds, victim vehicle speeds, and perception models.

cs.CR↗
arXiv2026

Adaptive Multi-Prior Lasso for High-Dimensional Generalized Linear Models

Incorporation of external information into high-dimensional modeling for gene expression data has been shown, both theoretically and empirically, to substantially enhance performance. Such external information, sometimes referred to as prior information or priors, has become increasingly accessible from multiple sources, yet its reliability may vary considerably. Existing approaches often integrate these priors without sufficiently accounting for their quality, which may result in unsatisfactory or even misleading results. To effectively and selectively exploit such priors, we propose adaptive Multi-Prior Lasso, a novel regularization approach that simultaneously identifies reliable prior sources and integrates them to improve model performance. For high-dimensional generalized linear models (GLMs), an adaptive data-driven weight is assigned to each prior, so that more reliable sources are emphasized while less credible ones are downweighted. Theoretical guarantees are established, and the proposed method is shown through extensive simulations to improve estimation, prediction, and variable selection. An application to TCGA breast cancer gene expression data further illustrates the practical value of the proposed method, showing that incorporating prior information from PubMed published studies improves model performance.

stat.ME↗
arXiv2026

Automatic Teller Machines for Offline E-cash

Electronic cash (e-cash) is a digital alternative to physical currency that allows anonymous transactions between users and merchants. Typically, coins in an e-cash scheme are only dispensed through a central bank. A drawback of this approach is that the bank is always on the critical path during withdrawals, and if a reliable connection to the bank is temporarily unavailable, users may be unable to withdraw coins in a timely fashion. As with physical currency, there are benefits to supporting a decentralized infrastructure where withdrawals can be performed without involving the bank in the critical path. We propose the design of a new cryptographic bearer token that can be dispensed by automatic teller machines (ATM) in a fully offline e-cash scheme. Such bearer tokens provide anonymity, unforgeability and untraceability, i.e., users cannot be tracked by their spending activities or the locations of withdrawal. We formalize the requirements of an e-cash scheme with multiple issuers and propose an efficient design building on top of the compact e-cash protocol of Camenisch et al. (EUROCRYPT 2005). Our construction leverages an unforgeable and doubly-anonymous voucher that allows a one-time transfer of coins between an ATM and a user, while hiding their identities from parties not involved in the transaction.

cs.CR↗
arXiv2025

Plato: Plan to Efficiently Decode for Large Language Model Inference

Large language models (LLMs) have achieved remarkable success in natural language tasks, but their inference incurs substantial computational and memory overhead. To improve efficiency, parallel decoding methods like Skeleton-of-Thought (SoT) decompose prompts into sub-problems for concurrent processing. However, these methods significantly compromise answer quality by treating semantically linked sub-problems as independent. We propose Plato, a novel approach that co-designs algorithms and systems for semantic-aware parallel decoding. Plato leverages LLMs to organize sub-problems into a dependency graph based on logical and causal relationships, enabling concurrent decoding of non-dependent nodes while preserving answer coherence and quality. To further enhance efficiency, Plato pipelines planning and node decoding stages, implements a global context cache, and carefully structures node inference prompts to maximize key-value cache reuse and minimize overhead. Our evaluations show that Plato improves throughput by 68% over autoregressive decoding while achieving a 40% net win rate in answer quality. Compared to SoT, Plato demonstrates a remarkable 90% quality net-win rate. Ablation studies reveal that our pipeline design improves speedup by 29%, while our KV cache reuse optimization reduces overhead by 75%.

cs.CL↗
arXiv2025

LLM Safeguard is a Double-Edged Sword: Exploiting False Positives for Denial-of-Service Attacks

Safety is a paramount concern for large language models (LLMs) in open deployment, motivating the development of safeguard methods that enforce ethical and responsible use through safety alignment or guardrail mechanisms. Jailbreak attacks that exploit the \emph{false negatives} of safeguard methods have emerged as a prominent research focus in the field of LLM security. However, we found that the malicious attackers could also exploit false positives of safeguards, i.e., fooling the safeguard model to block safe content mistakenly, leading to a denial-of-service (DoS) affecting LLM users. To bridge the knowledge gap of this overlooked threat, we explore multiple attack methods that include inserting a short adversarial prompt into user prompt templates and corrupting the LLM on the server by poisoned fine-tuning. In both ways, the attack triggers safeguard rejections of user requests from the client. Our evaluation demonstrates the severity of this threat across multiple scenarios. For instance, in the scenario of white-box adversarial prompt injection, the attacker can use our optimization process to automatically generate seemingly safe adversarial prompts, approximately only 30 characters long, that universally block over 97% of user requests on Llama Guard 3. These findings reveal a new dimension in LLM safeguard evaluation -- adversarial robustness to false positives.

cs.CR↗
arXiv2025

What Would Trojans Do? Exploiting Partial-Information Vulnerabilities in Autonomous Vehicle Sensing

Safety-critical sensors in autonomous vehicles (AVs) form an essential part of the vehicle's trusted computing base (TCB), yet they are highly susceptible to attacks. Alarmingly, Tier 1 manufacturers have already exposed vulnerabilities to attacks introducing Trojans that can stealthily alter sensor outputs. We analyze the feasible capability and safety-critical outcomes of an attack on sensing at a cyber level. To further address these threats, we design realistic attacks in AV simulators and real-world datasets under two practical constraints: attackers (1) possess only partial information and (2) are constrained by data structures that maintain sensor integrity.Examining the role of camera and LiDAR in multi-sensor AVs, we find that attacks targeting only the camera have minimal safety impact due to the sensor fusion system's strong reliance on 3D data from LiDAR. This reliance makes LiDAR-based attacks especially detrimental to safety. To mitigate the vulnerabilities, we introduce security-aware sensor fusion incorporating (1) a probabilistic data-asymmetry monitor and (2) a scalable track-to-track fusion of 3D LiDAR and monocular detections (T2T-3DLM). We demonstrate that these methods significantly diminish attack success rate.

cs.CR↗
arXiv2025

Compute Or Load KV Cache? Why Not Both?

Large Language Models (LLMs) are increasingly deployed in large-scale online services, enabling sophisticated applications. However, the computational overhead of generating key-value (KV) caches in the prefill stage presents a major bottleneck, particularly for long-context inputs. Prefix caching mitigates this issue by storing KV caches for reuse, reducing redundant computation. Despite its advantages, prefix caching suffers from high latency due to the limited I/O bandwidth of storage devices, constraining inference efficiency. To address this challenge, we introduce Cake, a novel KV cache loading system that optimally utilizes both computational and I/O resources in parallel. Cake employs a bidirectional scheduling strategy that dynamically balances KV cache computation and loading, ensuring efficient resource utilization. Additionally, Cake incorporates an adaptive scheduling mechanism that seamlessly integrates with non-prefix caching requests, improving system throughput and adapting to fluctuating resource availabilty. Through extensive evaluations across various hardware configurations, datasets, and storage conditions, Cake achieves on average 2.6x reduction in Time to First Token (TTFT) compared to compute-only and I/O-only methods. Our findings highlight Cake as an effective and practical solution for optimizing long-context LLM inference, bridging the gap between computation and I/O efficiency in large-scale AI deployments.

cs.LG↗
arXiv2024

Cocoon: Robust Multi-Modal Perception with Uncertainty-Aware Sensor Fusion

An important paradigm in 3D object detection is the use of multiple modalities to enhance accuracy in both normal and challenging conditions, particularly for long-tail scenarios. To address this, recent studies have explored two directions of adaptive approaches: MoE-based adaptive fusion, which struggles with uncertainties arising from distinct object configurations, and late fusion for output-level adaptive fusion, which relies on separate detection pipelines and limits comprehensive understanding. In this work, we introduce Cocoon, an object- and feature-level uncertainty-aware fusion framework. The key innovation lies in uncertainty quantification for heterogeneous representations, enabling fair comparison across modalities through the introduction of a feature aligner and a learnable surrogate ground truth, termed feature impression. We also define a training objective to ensure that their relationship provides a valid metric for uncertainty quantification. Cocoon consistently outperforms existing static and adaptive methods in both normal and challenging conditions, including those with natural and artificial corruptions. Furthermore, we show the validity and efficacy of our uncertainty metric across diverse datasets.

cs.CV↗
arXiv2023

Exploring the Limits of ChatGPT in Software Security Applications

Large language models (LLMs) have undergone rapid evolution and achieved remarkable results in recent times. OpenAI's ChatGPT, backed by GPT-3.5 or GPT-4, has gained instant popularity due to its strong capability across a wide range of tasks, including natural language tasks, coding, mathematics, and engaging conversations. However, the impacts and limits of such LLMs in system security domain are less explored. In this paper, we delve into the limits of LLMs (i.e., ChatGPT) in seven software security applications including vulnerability detection/repair, debugging, debloating, decompilation, patching, root cause analysis, symbolic execution, and fuzzing. Our exploration reveals that ChatGPT not only excels at generating code, which is the conventional application of language models, but also demonstrates strong capability in understanding user-provided commands in natural languages, reasoning about control and data flows within programs, generating complex data structures, and even decompiling assembly code. Notably, GPT-4 showcases significant improvements over GPT-3.5 in most security tasks. Also, certain limitations of ChatGPT in security-related tasks are identified, such as its constrained ability to process long code contexts.

cs.CR↗