One Single Hub Text Breaks CLIP: Identifying Vulnerabilities in Cross-Modal Encoders Via Hubness

The hubness problem, in which hub embeddings are close to many unrelated examples, occurs often in high-dimensional embedding spaces and may pose a practical threat for purposes such as information retrieval and automatic evaluation metrics. In particular, since cross-modal similarity between tex...

Enhancing Targeted Adversarial Attacks on Large Vision-Language Models through Intermediate Projector Guidance

Targeted adversarial attacks are essential for proactively identifying security flaws in Vision-Language Models before real-world deployment. However, current methods perturb images to maximize global similarity with the target text or reference image at the encoder level, collapsing rich visual...

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...

Building Trustworthy Multimodal AI: a Review of Fairness, Transparency, and Ethics in Vision-Language Tasks

Objective: This review explores the trustworthiness of multimodal artificial intelligence AI systems, specifically focusing on vision-language tasks. It addresses critical challenges related to fairness, transparency, and ethical implications in these systems, providing a comparative analysis of...

gstreamer1-plugins-good: OOB-write in convert_to_s334_1a

A flaw was found in the GStreamer library. An out-of-bounds write in the MP4/MOV demuxer when handling CEA608 Closed Caption tracks can lead to crashes for certain input files. This vulnerability allows a malicious third party to trigger a crash of the application and perform code execution throu...

gstreamer1-plugins-good: OOB-write in convert_to_s334_1a

A flaw was found in the GStreamer library. An out-of-bounds write in the MP4/MOV demuxer when handling CEA608 Closed Caption tracks can lead to crashes for certain input files. This vulnerability allows a malicious third party to trigger a crash of the application and perform code execution throu...

Accessibility and usability for all in Azure Sentinel

As a father of a child on the Autism spectrum who relies completely on digital media for his learning, I fully appreciate the impact that digital accessibility can have on people with disabilities. Designing with accessibility in mind greatly expands the impact of Microsoft solutions. What many...

CVE-2012-0004

Unspecified vulnerability in DirectShow in DirectX in Microsoft Windows XP SP2 and SP3, Windows Server 2003 SP2, Windows Vista SP2, Windows Server 2008 SP2, R2, and R2 SP1, and Windows 7 Gold and SP1 allows remote attackers to execute arbitrary code via a crafted media file, related to Quartz.dll...

Fedora Update for zvbi FEDORA-2007-0175

Check for the Version of zvbi OpenVAS Vulnerability Test Fedora Update for zvbi FEDORA-2007-0175 Authors: System Generated Check Copyright: Copyright c 2009 Greenbone Networks GmbH, http://www.greenbone.net This program is free software; you can redistribute it and/or modify it under the terms of...

[SECURITY] Fedora 7 Update: zvbi-0.2.25-1.fc7

ZVBI provides functions to capture and decode VBI data. The vertical blanki ng interval VBI is an interval in a television signal that temporarily suspe nds transmission of the signal for the electron gun to move back up to the first line of the television screen to trace the next screen field. T...