A Comparative Study of 3D Person Detection: Sensor Modalities and Robustness in Diverse Indoor and Outdoor Environments

Accurate 3D person detection is critical for safety in applications such as robotics, industrial monitoring, and surveillance. This work presents a systematic evaluation of 3D person detection using camera-only, LiDAR-only, and camera-LiDAR fusion. While most existing research focuses on autonomous driving, we explore detection performance and robustness in diverse indoor and outdoor scenes using the JRDB dataset. We compare three representative models - BEVDepth (camera), PointPillars (LiDAR), and DAL (camera-LiDAR fusion) - and analyze their behavior under varying occlusion and distance levels. Our results show that the fusion-based approach consistently outperforms single-modality models, particularly in challenging scenarios. We further investigate robustness against sensor corruptions and misalignments, revealing that while DAL offers improved resilience, it remains sensitive to sensor misalignment and certain LiDAR-based corruptions. In contrast, the camera-based BEVDepth model showed the lowest performance and was most affected by occlusion, distance, and noise. Our findings highlight the importance of utilizing sensor fusion for enhanced 3D person detection, while also underscoring the need for ongoing research to address the vulnerabilities inherent in these systems.

Detecting Misbehaviors of Large Vision-Language Models by Evidential Uncertainty Quantification

Large vision-language models (LVLMs) have shown substantial advances in multimodal understanding and generation. However, when presented with incompetent or adversarial inputs, they frequently produce unreliable or even harmful content, such as fact hallucinations or dangerous instructions. This misalignment with human expectations, referred to as \emph{misbehaviors} of LVLMs, raises serious concerns for deployment in critical applications. These misbehaviors are found to stem from epistemic uncertainty, specifically either conflicting internal knowledge or the absence of supporting information. However, existing uncertainty quantification methods, which typically capture only overall epistemic uncertainty, have shown limited effectiveness in identifying such issues. To address this gap, we propose Evidential Uncertainty Quantification (EUQ), a fine-grained method that captures both information conflict and ignorance for effective detection of LVLM misbehaviors. In particular, we interpret features from the model output head as either supporting (positive) or opposing (negative) evidence. Leveraging Evidence Theory, we model and aggregate this evidence to quantify internal conflict and knowledge gaps within a single forward pass. We extensively evaluate our method across four categories of misbehavior, including hallucinations, jailbreaks, adversarial vulnerabilities, and out-of-distribution (OOD) failures, using state-of-the-art LVLMs, and find that EUQ consistently outperforms strong baselines, showing that hallucinations correspond to high internal conflict and OOD failures to high ignorance. Furthermore, layer-wise evidential uncertainty dynamics analysis helps interpret the evolution of internal representations from a new perspective. The source code is available at https://github.com/HT86159/EUQ.

CASTLE: A Comprehensive Benchmark for Evaluating Student-Tailored Personalized Safety in Large Language Models

Large language models (LLMs) have advanced the development of personalized learning in education. However, their inherent generation mechanisms often produce homogeneous responses to identical prompts. This one-size-fits-all mechanism overlooks the substantial heterogeneity in students cognitive and psychological, thereby posing potential safety risks to vulnerable groups. Existing safety evaluations primarily rely on context-independent metrics such as factual accuracy, bias, or toxicity, which fail to capture the divergent harms that the same response might cause across different student attributes. To address this gap, we propose the concept of Student-Tailored Personalized Safety and construct CASTLE based on educational theories. This benchmark covers 15 educational safety risks and 14 student attributes, comprising 92,908 bilingual scenarios. We further design three evaluation metrics: Risk Sensitivity, measuring the model ability to detect risks; Emotional Empathy, evaluating the model capacity to recognize student states; and Student Alignment, assessing the match between model responses and student attributes. Experiments on 18 SOTA LLMs demonstrate that CASTLE poses a significant challenge: all models scored below an average safety rating of 2.3 out of 5, indicating substantial deficiencies in personalized safety assurance.

GNSS SpAmming: a spoofing-based GNSS denial-of-service attack

GNSSs are vulnerable to attacks of two kinds: jamming (i.e. denying access to the signal) and spoofing (i.e. impersonating a legitimate satellite). These attacks have been extensively studied, and we have a myriad of countermeasures to mitigate them. In this paper we expose a new type of attack: SpAmming, which combines both approaches to achieve the same effects in a more subtle way. Exploiting the CDMA multiplexing present in most GNSSs, and through a spoofing attack, this approach leads the receiver to lose access to the signal of a legitimate satellite, which would be equivalent to a denial of service; but in this case the existing countermeasures against jamming or spoofing would not allow safeguarding its effectiveness, as it is neither of them. An experimental proof-of-concept is presented in which its impact is evaluated as a function of the previous state of the receiver. Using an SDR-based system developed at the Space Security Centre, the attack is executed against a cold-started receiver, a warm-started receiver, and a receiver that has already acquired the PVT solution and is navigating. Different attack configurations are also tested, starting from a raw emission of the false signal, to surgical Doppler effect configuration, code offset, etc. Although it is shown to be particularly successful against cold-started receivers, the results show that it is also effective in other scenarios, especially if accompanied by other attacks. We will conclude the article by outlining possible countermeasures to detect and, eventually, counteract it; and possible avenues of research to better understand its impact, especially for authenticated services such as OSNMA, and to characterize it in order to improve the response to similar attacks.

Trojan Attacks on Neural Network Controllers for Robotic Systems

Neural network controllers are increasingly deployed in robotic systems for tasks such as trajectory tracking and pose stabilization. However, their reliance on potentially untrusted training pipelines or supply chains introduces significant security vulnerabilities. This paper investigates backdoor (Trojan) attacks against neural controllers, using a differential-drive mobile robot platform as a case study. In particular, assuming that the robot's tracking controller is implemented as a neural network, we design a lightweight, parallel Trojan network that can be embedded within the controller. This malicious module remains dormant during normal operation but, upon detecting a highly specific trigger condition defined by the robot's pose and goal parameters, compromises the primary controller's wheel velocity commands, resulting in undesired and potentially unsafe robot behaviours. We provide a proof-of-concept implementation of the proposed Trojan network, which is validated through simulation under two different attack scenarios. The results confirm the effectiveness of the proposed attack and demonstrate that neural network-based robotic control systems are subject to potentially critical security threats.

From Data to Behavior: Predicting Unintended Model Behaviors Before Training

Large Language Models (LLMs) can acquire unintended biases from seemingly benign training data even without explicit cues or malicious content. Existing methods struggle to detect such risks before fine-tuning, making post hoc evaluation costly and inefficient. To address this challenge, we introduce Data2Behavior, a new task for predicting unintended model behaviors prior to training. We also propose Manipulating Data Features (MDF), a lightweight approach that summarizes candidate data through their mean representations and injects them into the forward pass of a base model, allowing latent statistical signals in the data to shape model activations and reveal potential biases and safety risks without updating any parameters. MDF achieves reliable prediction while consuming only about 20% of the GPU resources required for fine-tuning. Experiments on Qwen3-14B, Qwen2.5-32B-Instruct, and Gemma-3-12b-it confirm that MDF can anticipate unintended behaviors and provide insight into pre-training vulnerabilities.

COMET: Codebook-based Online-adaptive Multi-scale Embedding for Time-series Anomaly Detection

Time series anomaly detection is a critical task across various industrial domains. However, capturing temporal dependencies and multivariate correlations within patch-level representation learning remains underexplored, and reliance on single-scale patterns limits the detection of anomalies across different temporal ranges. Furthermore, focusing on normal data representations makes models vulnerable to distribution shifts at inference time. To address these limitations, we propose Codebook-based Online-adaptive Multi-scale Embedding for Time-series anomaly detection (COMET), which consists of three key components: (1) Multi-scale Patch Encoding captures temporal dependencies and inter-variable correlations across multiple patch scales. (2) Vector-Quantized Coreset learns representative normal patterns via codebook and detects anomalies with a dual-score combining quantization error and memory distance. (3) Online Codebook Adaptation generates pseudo-labels based on codebook entries and dynamically adapts the model at inference through contrastive learning. Experiments on five benchmark datasets demonstrate that COMET achieves the best performance in 36 out of 45 evaluation metrics, validating its effectiveness across diverse environments.

Co-RedTeam: Orchestrated Security Discovery and Exploitation with LLM Agents

Large language models (LLMs) have shown promise in assisting cybersecurity tasks, yet existing approaches struggle with automatic vulnerability discovery and exploitation due to limited interaction, weak execution grounding, and a lack of experience reuse. We propose Co-RedTeam, a security-aware multi-agent framework designed to mirror real-world red-teaming workflows by integrating security-domain knowledge, code-aware analysis, execution-grounded iterative reasoning, and long-term memory. Co-RedTeam decomposes vulnerability analysis into coordinated discovery and exploitation stages, enabling agents to plan, execute, validate, and refine actions based on real execution feedback while learning from prior trajectories. Extensive evaluations on challenging security benchmarks demonstrate that Co-RedTeam consistently outperforms strong baselines across diverse backbone models, achieving over 60% success rate in vulnerability exploitation and over 10% absolute improvement in vulnerability detection. Ablation and iteration studies further confirm the critical role of execution feedback, structured interaction, and memory for building robust and generalizable cybersecurity agents.

Semantic-level Backdoor Attack against Text-to-Image Diffusion Models

Text-to-image (T2I) diffusion models are widely adopted for their strong generative capabilities, yet remain vulnerable to backdoor attacks. Existing attacks typically rely on fixed textual triggers and single-entity backdoor targets, making them highly susceptible to enumeration-based input defenses and attention-consistency detection. In this work, we propose Semantic-level Backdoor Attack (SemBD), which implants backdoors at the representation level by defining triggers as continuous semantic regions rather than discrete textual patterns. Concretely, SemBD injects semantic backdoors by distillation-based editing of the key and value projection matrices in cross-attention layers, enabling diverse prompts with identical semantic compositions to reliably activate the backdoor attack. To further enhance stealthiness, SemBD incorporates a semantic regularization to prevent unintended activation under incomplete semantics, as well as multi-entity backdoor targets that avoid highly consistent cross-attention patterns. Extensive experiments demonstrate that SemBD achieves a 100% attack success rate while maintaining strong robustness against state-of-the-art input-level defenses.

Adversarial Reward Auditing for Active Detection and Mitigation of Reward Hacking

Reinforcement Learning from Human Feedback (RLHF) remains vulnerable to reward hacking, where models exploit spurious correlations in learned reward models to achieve high scores while violating human intent. Existing mitigations rely on static defenses that cannot adapt to novel exploitation strategies. We propose Adversarial Reward Auditing (ARA), a framework that reconceptualizes reward hacking as a dynamic, competitive game. ARA operates in two stages: first, a Hacker policy discovers reward model vulnerabilities while an Auditor learns to detect exploitation from latent representations; second, Auditor-Guided RLHF (AG-RLHF) gates reward signals to penalize detected hacking, transforming reward hacking from an unobservable failure into a measurable, controllable signal. Experiments across three hacking scenarios demonstrate that ARA achieves the best alignment-utility tradeoff among all baselines: reducing sycophancy to near-SFT levels while improving helpfulness, decreasing verbosity while achieving the highest ROUGE-L, and suppressing code gaming while improving Pass@1. Beyond single-domain evaluation, we show that reward hacking, detection, and mitigation all generalize across domains -- a Hacker trained on code gaming exhibits increased sycophancy despite no reward for this behavior, and an Auditor trained on one domain effectively suppresses exploitation in others, enabling efficient multi-domain defense with a single model.