CVE-2026-75837

Grav before 2.0.14 fails to guard the access field in the core group blueprint with the required security@: admin.super restriction. A delegated admin.users operator can save a group with accessadminsuper=true to escalate to super-admin, gaining scheduler and Twig evaluation capabilities...

CVE-2026-75837 Grav before 2.0.14 Privilege Escalation via Group Access Field

Grav before 2.0.14 fails to guard the access field in the core group blueprint with the required security@: admin.super restriction. A delegated admin.users operator can save a group with accessadminsuper=true to escalate to super-admin, gaining scheduler and Twig evaluation capabilities...

CVE-2026-19608

A flaw in the group policy provider of Keycloak authorization services allows unauthorized access to protected resources. The root cause is that group-based policy evaluation uses tokens containing only group names rather than full paths . When two groups in different parts of the organization hi...

CVE-2026-19608

A flaw was found in the group policy provider of Keycloak authorization services, which is used to manage fine-grained access control to resources. The issue occurs when the system evaluates group-based policies using tokens that only contain group names rather than full paths. If two groups in...

EUVD-2026-60581

A flaw was found in the group policy provider of Keycloak authorization services, which is used to manage fine-grained access control to resources. The issue occurs when the system evaluates group-based policies using tokens that only contain group names rather than full paths. If two groups in...

CVE-2026-19608

A flaw was found in the group policy provider of Keycloak authorization services, which is used to manage fine-grained access control to resources. The issue occurs when the system evaluates group-based policies using tokens that only contain group names rather than full paths. If two groups in...

CVE-2024-58375

A flaw was found in OpenTofu. When users enable static evaluation of module sources, versions, and backend configurations, the software does not properly restrict sensitive variables and locals. This oversight can lead to the exposure of values intended to be sensitive through these configuration...

LLMs and Contextual Integrity

I have been thinking a lot about AI and integrity. Part of that is contextual integrity. I recently found two papers on the topic. "CIMemories: A Compositional Benchmark for Contextual Integrity of Persistent Memory in LLMs": Abstract: Large Language Models LLMs increasingly use persistent memory...

PT-2026-76916

A flaw was found in the group policy provider of Keycloak authorization services, which is used to manage fine-grained access control to resources. The issue occurs when the system evaluates group-based policies using tokens that only contain group names rather than full paths. If two groups in...

CVE-2026-75837: Improper Privilege Management

Grav before 2.0.14 fails to guard the access field in the core group blueprint with the required security@: admin.super restriction. A delegated admin.users operator can save a group with accessadminsuper=true to escalate to super-admin, gaining scheduler and Twig evaluation capabilities...

Fair ASR: Re-Evaluating Black-Box Jailbreaks under Shared Target-Call Budgets

Reliable jailbreak evaluation is essential for assessing LLM safety, but most existing studies rely solely on attack success rate ASR without accounting for its dependence on attack budgets, resulting in unfair comparisons across methods. Existing compute-aware evaluations reduce heterogeneous...

AI-Infra-Guard v4.5.2

📖 Documentation | 🌐 🇨🇳 中文 · 🇯🇵 日本語 · 🇪🇸 Español · 🇩🇪 Deutsch · 🇫🇷 Français · 🇰🇷 한국어 · 🇧🇷 Português · 🇷🇺 Русский 🚀 AI Red Teaming Platform by Tencent Zhuque Lab A.I.G AI-Infra-Guard integrates capabilities such as ClawScanOpenClaw Security Scan, Agent Scan，AI infra vulnerability scan, MCP Server &...

EUVD-2026-60355

Cross Site Scripting vulnerability in Squirro Cognitive Search 3.14.2 allows a remote attacker to execute arbtirary code via the Email Notification, Create Evaluation Sets and HTML Editor functions...

CVE-2026-50771

Cross Site Scripting vulnerability in Squirro Cognitive Search 3.14.2 allows a remote attacker to execute arbtirary code via the Email Notification, Create Evaluation Sets and HTML Editor functions...

CVE-2026-50771

Cross Site Scripting vulnerability in Squirro Cognitive Search 3.14.2 allows a remote attacker to execute arbtirary code via the Email Notification, Create Evaluation Sets and HTML Editor functions...

Security of Foundation-Model-Powered Embodied Agents: Attack Surfaces, Attacks, Defenses, and Evaluation

Foundation models are increasingly used for perception, reasoning, planning, and action generation in embodied agents, creating security risks that can propagate from digital inputs to physical behavior. Existing surveys often organize threats by mechanisms such as jailbreaks, prompt injection,...

CVE-2026-50771: Improper Neutralization of Input During Web Page Generation

Cross Site Scripting vulnerability in Squirro Cognitive Search 3.14.2 allows a remote attacker to execute arbtirary code via the Email Notification, Create Evaluation Sets and HTML Editor functions...

CVE-2026-50771

Cross Site Scripting vulnerability in Squirro Cognitive Search 3.14.2 allows a remote attacker to execute arbtirary code via the Email Notification, Create Evaluation Sets and HTML Editor functions...

CVE-2024-58375

OpenTofu versions 1.8.0 through 1.8.2 do not properly restrict sensitive variables and locals when users have opted into static evaluation of module sources, versions, and backend configurations. As a result, values marked as sensitive may be exposed through these configuration elements instead o...

CVE-2026-74785 Scriban before 7.0.0 Denial of Service via Unbounded Resource Consumption

Scriban before 7.0.0 contains three distinct denial-of-service vulnerabilities in expression evaluation that bypass existing safety controls through unbounded string multiplication, uncontrolled BigInteger shift operations, and LoopLimit bypass via range enumeration in builtin functions. Attacker...