GHSA-QW65-CVWX-89V3 vulnerabilities

Vulnerabilities for packages: serve, prism, yaml-language-server, opensearch-dashboards, vitess, renovate, jupyter-base-notebook...

CVE-2026-84292 vulnerabilities

Vulnerabilities for packages: serve, prism, yaml-language-server, opensearch-dashboards, vitess, renovate, jupyter-base-notebook...

CVE-2026-102830

A flaw was found in JupyterLab. Incomplete validation of language pack headers allows arbitrary JavaScript code to be injected and executed during string translation. An attacker could exploit this flaw by enticing a user to load a malicious third-party language pack, resulting in code execution...

CVE-2026-102830 JupyterLab: Cross-site scripting (XSS) in JupyterLab via crafted language package (jupyterlab.json)

JupyterLab is an extensible environment for interactive and reproducible computing, based on the Jupyter Notebook Architecture. From JupyterLab 3.0.0 until 4.5.11 and 4.6.4, and in JupyterLite Core 0.8.3 and earlier, the Plural-Forms header in a selected third-party language pack can append...

CVE-2026-102830

A vulnerability in JupyterLab (versions 3.0.0 through 4.5.11 and 4.6.4) and JupyterLite Core (0.8.3 and earlier) allows for the execution of arbitrary JavaScript. The root cause is a prefix-only regular-expression validation in the Plural-Forms header of certain third-party language packs, which ...

CVE-2026-102570

ClipBucket v5 through 5.5.3-197 contains a time-based blind SQL injection vulnerability in the language update function where the languageid parameter is concatenated unescaped into the WHERE clause of an UPDATE statement. An authenticated administrator with basicsettings permission can inject...

EUVD-2026-88971

ClipBucket v5 through 5.5.3-197 contains a time-based blind SQL injection vulnerability in the language update function where the languageid parameter is concatenated unescaped into the WHERE clause of an UPDATE statement. An authenticated administrator with basicsettings permission can inject...

PT-2026-102628

ClipBucket v5 through 5.5.3-197 contains a time-based blind SQL injection vulnerability in the language update function where the language id parameter is concatenated unescaped into the WHERE clause of an UPDATE statement. An authenticated administrator with basic settings permission can inject...

Selective Channel Restoration for Backdoored Vision-Language Models

Vision-language models VLMs exhibit strong multimodal capabilities but remain vulnerable to backdoors implanted through poisoned fine-tuning data. Existing defenses often require extensive parameter updates during fine-tuning or incur per-query overhead during inference. To address these...

Security-Enhanced Seed-Based Weight Quantization for Large Language Models

Large language models LLMs incur substantial storage, memory-bandwidth and energy costs, motivating compact weight representations. Existing seed-based compression methods reconstruct weights from compact pseudo-random representations but do not explicitly account for the non-uniform sensitivity ...

The Geometry of Harmfulness in Multi-Turn Attacks

Large language models LLMs remain vulnerable to adversarial attacks that circumvent safety alignment to elicit harmful outputs. It remains unclear how harmfulness and refusal representations evolve over the course of multi-turn attacks, and why single-turn defenses are less effective in multi-tur...

Controlled Decoding Attacks on Black-Box LLMs

Manipulating next-token probabilities during generation can bypass the safety alignment of large language models. Existing approaches, however, rely on access to model weights or numerical token probabilities and therefore do not apply to interfaces that return only sampled text. Reconstructing...

Where Do LLMs Decide to Break the Rules? Mechanistic Localization of Prompt Injection Compliance

When a prompt injection attack succeeds, a Large Language Model LLM abandons its assigned system role to comply with an adversarial instruction. While prior work has extensively quantified how often this occurs, we ask a more fundamental question: where inside the network does the model actually...

Removing the NEEDLE in the Haystack: Backdoor Removal in LLMs Via Weight Orthogonalisation

Backdoor attacks can be implanted in Large Language Models LLMs during training, causing unwanted behaviour when a trigger appears in the input. Existing backdoor defences for LLMs attempt to remove the backdoor but inadvertently shift the model's output distribution to benign prompts, which can...

GO-2026-6539 nginx ignition has ParseAcceptLanguage `_` separator bypass that enables ~75x CPU amplification via Accept-Language header in i18nMiddleware in github.com/lucasdillmann/nginx-ignition

nginx ignition has ParseAcceptLanguage separator bypass that enables 75x CPU amplification via Accept-Language header in i18nMiddleware in github.com/lucasdillmann/nginx-ignition...

CVE-2026-54675: Improper Limitation of a Pathname to a Restricted Directory

FreePBX is an open source IP PBX. Prior to versions 16.0.10 and 17.0.5, a critical vulnerability exists in the sound language upload and conversion functionality that allows an authenticated attacker to perform arbitrary file writes, leading directly to remote code execution RCE. Authentication...

SEAD: A State-Based Perspective on Attack and Defense in Tool-Using Agents

Language-model agents increasingly use tools to act on external systems. Earlier actions can alter files, permissions, database records, or other state, making a later routine-looking action harmful. Yet the visible interaction may not reveal the underlying state needed to assess that action. We...

FinRT: Distilling Adaptive Red-Teaming Strategies into Reusable Adversarial Generators in Consumer Finance

In regulated industries like consumer finance, seemingly harmless user queries can exploit large language model vulnerabilities, triggering safety failures and pushing responses dangerously close to policy limits. Existing automated red-teaming methods trade off attack effectiveness against...

PT-2026-99959

FreePBX is an open source IP PBX. Prior to versions 16.0.10 and 17.0.5, a critical vulnerability exists in the sound language upload and conversion functionality that allows an authenticated attacker to perform arbitrary file writes, leading directly to remote code execution RCE. Authentication...

Distillation Defenses Easily Break after Reinforcement Learning

Distillation attacks copy the reasoning capabilities of closed-source large language models, allowing bad actors to replicate state-of-the-art performance at low cost. Attackers systematically collect a large volume of frontier model reasoning traces and then train i.e., "distill" their own model...