DAVSP: Safety Alignment for Large Vision-Language Models Via Deep Aligned Visual Safety Prompt

Large Vision-Language Models LVLMs have achieved impressive progress across various applications but remain vulnerable to malicious queries that exploit the visual modality. Existing alignment approaches typically fail to resist malicious queries while preserving utility on benign ones effectivel...

SQL Injection Vulnerability in Multimedia Integrated Service Display System of Beijing Shenzhou Vision Han Technology Co., Ltd (CNVD-C-2025-319811)

Ltd. is a deep-rooted enterprise in the field of visualization. A SQL injection vulnerability exists in the multimedia integrated business display system of Beijing Divine Vision Han Technology Co. Ltd, which can be exploited by attackers to obtain sensitive information from the database...

Attacking Attention of Foundation Models Disrupts Downstream Tasks

Foundation models represent the most prominent and recent paradigm shift in artificial intelligence. Foundation models are large models, trained on broad data that deliver high accuracy in many downstream tasks, often without fine-tuning. For this reason, models such as CLIP , DINO or Vision...

SoK: Data Reconstruction Attacks against Machine Learning Models: Definition, Metrics, and Benchmark

Data reconstruction attacks, which aim to recover the training dataset of a target model with limited access, have gained increasing attention in recent years. However, there is currently no consensus on a formal definition of data reconstruction attacks or appropriate evaluation metrics for...

Backdoor Attack on Vision Language Models with Stealthy Semantic Manipulation

Vision Language Models VLMs have shown remarkable performance, but are also vulnerable to backdoor attacks whereby the adversary can manipulate the model's outputs through hidden triggers. Prior attacks primarily rely on single-modality triggers, leaving the crucial cross-modal fusion nature of...

VLMs Can Aggregate Scattered Training Patches

Whitepaper called VLMs Can Aggregate Scattered Training Patches...

CVE-2024-53015 Use After Free in Computer Vision

Memory corruption while processing IOCTL command to handle buffers associated with a session...

CVE-2024-53015 Use After Free in Computer Vision

Memory corruption while processing IOCTL command to handle buffers associated with a session...

CVE-2024-53015

CVE-2024-53015 describes a memory corruption issue in Qualcomm chipsets triggered when processing IOCTL commands to handle buffers for a session. The vulnerability affects IOCTL buffer handling code and is evidenced by multiple feeds (NVD and Red Hat advisories) reporting memory corruption withou...

PT-2025-44855

Name of the Vulnerable Software and Affected Versions Safari versions prior to 26 tvOS versions prior to 26 watchOS versions prior to 26 iOS versions prior to 26 iPadOS versions prior to 26 visionOS versions prior to 26 Description A flaw exists due to improper memory handling when processing...

CVE-2025-48381

Computer Vision Annotation Tool CVAT is an interactive video and image annotation tool for computer vision. In versions starting from 2.4.0 to before 2.38.0, an authenticated CVAT user may be able to retrieve the IDs and names of all tasks, projects, labels, and the IDs of all jobs and quality...

CVE-2025-48381 CVAT has information disclosure via browsable API

Computer Vision Annotation Tool CVAT is an interactive video and image annotation tool for computer vision. In versions starting from 2.4.0 to before 2.38.0, an authenticated CVAT user may be able to retrieve the IDs and names of all tasks, projects, labels, and the IDs of all jobs and quality...

Computer Vision Annotation Tool 安全漏洞

Computer Vision Annotation Tool CVAT is a cvat.ai open source interactive video and image annotation tool for computer vision. A security vulnerability exists in Computer Vision Annotation Tool versions prior to 2.4.0 through 2.38.0, which stems from the possibility that an authenticated user may...

Disrupting Vision-Language Model-Driven Navigation Services Via Adversarial Object Fusion

We present Adversarial Object Fusion AdvOF, a novel attack framework targeting vision-and-language navigation VLN agents in service-oriented environments by generating adversarial 3D objects. While foundational models like Large Language Models LLMs and Vision Language Models VLMs have enhanced...

Spa-VLM: Stealthy Poisoning Attacks on RAG-Based VLM

With the rapid development of the Vision-Language Model VLM, significant progress has been made in Visual Question Answering VQA tasks. However, existing VLM often generate inaccurate answers due to a lack of up-to-date knowledge. To address this issue, recent research has introduced...

One Surrogate to Fool Them All: Universal, Transferable, and Targeted Adversarial Attacks with CLIP

Deep Neural Networks DNNs have achieved widespread success yet remain prone to adversarial attacks. Typically, such attacks either involve frequent queries to the target model or rely on surrogate models closely mirroring the target model -- often trained with subsets of the target model's traini...

TeleSparse: Practical Privacy-Preserving Verification of Deep Neural Networks

Verification of the integrity of deep learning inference is crucial for understanding whether a model is being applied correctly. However, such verification typically requires access to model weights and potentially sensitive or private training data. So-called Zero-knowledge Succinct...

SQL Injection Vulnerability in Multimedia Integrated Service Display System of Beijing Divine Vision Han Technology Co.

Ltd. is a deep-rooted enterprise in the field of visualization. A SQL injection vulnerability exists in the multimedia integrated business display system of Beijing Divine Vision Han Technology Co. Ltd, which can be exploited by attackers to obtain sensitive information from the database...

Impressions from PHDays Fest

Impressions from PHDays Fest. The scale was just insane. You walk and walk - and there's action everywhere, and all of it is PHDays, every bit of it. It totally blew my mind, I saw just a tiny fraction of everything that was going on. In the public area, I was impressed by the university pavilion...

CVE-2024-5474

A potential information disclosure vulnerability was reported in Lenovo's packaging of Dolby Vision Provisioning software prior to version 2.0.0.2 that could allow a local attacker to read files on the system with elevated privileges during installation of the package. Previously installed versio...