SoK: The Attack Surface of Agentic AI -- Tools, and Autonomy

Recent AI systems combine large language models with tools, external knowledge via retrieval-augmented generation RAG, and even autonomous multi-agent decision loops. This agentic AI paradigm greatly expands capabilities - but also vastly enlarges the attack surface. In this systematization, we m...

We Found Eight Attack Vectors Inside AWS Bedrock. Here's What Attackers Can Do with Them

AWS Bedrock is Amazon's platform for building AI-powered applications. It gives developers access to foundation models and the tools to connect those models directly to enterprise data and systems. That connectivity is what makes it powerful – but it’s also what makes Bedrock a target. When an AI...

Model Context Protocol Threat Modeling and Analyzing Vulnerabilities to Prompt Injection with Tool Poisoning

The Model Context Protocol MCP has rapidly emerged as a universal standard for connecting AI assistants to external tools and data sources. While MCP simplifies integration between AI applications and various services, it introduces significant security vulnerabilities, particularly on the client...

Towards Secure Retrieval-Augmented Generation: A Comprehensive Review of Threats, Defenses and Benchmarks

Retrieval-Augmented Generation RAG significantly mitigates the hallucinations and domain knowledge deficiency in large language models by incorporating external knowledge bases. However, the multi-module architecture of RAG introduces complex system-level security vulnerabilities. Guided by the R...

BIT-PARSE-2026-32878 Parse Server vulnerable to schema poisoning via prototype pollution in deep copy

Parse Server is an open source backend that can be deployed to any infrastructure that can run Node.js. Prior to 9.6.0 and 8.6.44, an attacker can bypass the default request keyword denylist protection and the class-level permission for adding fields by sending a crafted request that exploits...

Memory Poisoning and Secure Multi-Agent Systems

Memory poisoning attacks for Agentic AI and multi-agent systems MAS have recently caught attention. It is partially due to the fact that Large Language Models LLMs facilitate the construction and deployment of agents. Different memory systems are being used nowadays in this context, including...

GHSA-VFX2-HV2G-XJ5F Protocol-Relative URL Injection via Single Backslash Bypass in Angular SSR

An Open Redirect vulnerability exists in @angular/ssr due to an incomplete fix for CVE-2026-27738. While the original fix successfully blocked multiple leading slashes e.g., ///, the internal validation logic fails to account for a single backslash \ bypass. When an Angular SSR application is...

Protocol-Relative URL Injection via Single Backslash Bypass in Angular SSR

An Open Redirect vulnerability exists in @angular/ssr due to an incomplete fix for CVE-2026-27738. While the original fix successfully blocked multiple leading slashes e.g., ///, the internal validation logic fails to account for a single backslash \ bypass. When an Angular SSR application is...

CVE-2026-32622

SQLBot (versions ≤ 1.5.x) exposes a Stored Prompt Injection vulnerability consisting of three chained flaws: (1) missing permission check on the Excel upload API allowing any authenticated user to upload malicious terminology, (2) unsanitized storage of terminology descriptions containing dangero...

Juju has unauthorized update of out-of-scope Vault secrets

An authorization bypass vulnerability in the Vault secrets back-end implementation of Juju versions 3.1.6 through 3.6.18 allows an authenticated unit agent to perform unauthorized updates to secret revisions. With sufficient information, an attacker can poison any existing secret revision within...

Ubuntu 16.04 LTS / 18.04 LTS / 20.04 LTS / 22.04 LTS / 24.04 LTS : Bouncy Castle vulnerabilities (USN-8108-1)

The remote Ubuntu 16.04 LTS / 18.04 LTS / 20.04 LTS / 22.04 LTS / 24.04 LTS host has packages installed that are affected by multiple vulnerabilities as referenced in the USN-8108-1 advisory. It was discovered that Bouncy Castle did not sanitize user input when inserting it into an LDAP search...

A Framework for Formalizing LLM Agent Security

Security in LLM agents is inherently contextual. For example, the same action taken by an agent may represent legitimate behavior or a security violation depending on whose instruction led to the action, what objective is being pursued, and whether the action serves that objective. However,...

CVE-2026-32878 Parse Server vulnerable to schema poisoning via prototype pollution in deep copy

Parse Server is an open source backend that can be deployed to any infrastructure that can run Node.js. Prior to 9.6.0-alpha.20 and 8.6.44, an attacker can bypass the default request keyword denylist protection and the class-level permission for adding fields by sending a crafted request that...

CVE-2026-32878

Parse Server is vulnerable to prototype pollution in its deep copy path prior to versions 9.6.0-alpha.20 and 8.6.44. An attacker can bypass the default denylist and class-level field-adding permissions by crafting a request, allowing injection of fields into locked schemas and causing permanent s...

CVE-2026-32878 Parse Server vulnerable to schema poisoning via prototype pollution in deep copy

Parse Server is an open source backend that can be deployed to any infrastructure that can run Node.js. Prior to 9.6.0-alpha.20 and 8.6.44, an attacker can bypass the default request keyword denylist protection and the class-level permission for adding fields by sending a crafted request that...

CVE-2026-32878 Parse Server vulnerable to schema poisoning via prototype pollution in deep copy

Parse Server is an open source backend that can be deployed to any infrastructure that can run Node.js. Prior to 9.6.0-alpha.20 and 8.6.44, an attacker can bypass the default request keyword denylist protection and the class-level permission for adding fields by sending a crafted request that...

USN-8108-1: Bouncy Castle vulnerabilities

It was discovered that Bouncy Castle did not sanitize user input when inserting it into an LDAP search filter. An attacker could possibly use this issue to perform an LDAP injection attack. This issue only affected Ubuntu 16.04 LTS, Ubuntu 18.04 LTS, Ubuntu 20.04 LTS, and Ubuntu 22.04 LTS...

USN-8108-1 bouncycastle vulnerabilities

It was discovered that Bouncy Castle did not sanitize user input when inserting it into an LDAP search filter. An attacker could possibly use this issue to perform an LDAP injection attack. This issue only affected Ubuntu 16.04 LTS, Ubuntu 18.04 LTS, Ubuntu 20.04 LTS, and Ubuntu 22.04 LTS...

undertow-core: Undertow HTTP Server Fails to Reject Malformed Host Headers Leading to Potential Cache Poisoning and SSRF

A flaw was found in the Undertow HTTP server core, which is used in WildFly, JBoss EAP, and other Java applications. The Undertow library fails to properly validate the Host header in incoming HTTP requests.As a result, requests containing malformed or malicious Host headers are processed without...

undertow-core: Undertow HTTP Server Fails to Reject Malformed Host Headers Leading to Potential Cache Poisoning and SSRF

A flaw was found in the Undertow HTTP server core, which is used in WildFly, JBoss EAP, and other Java applications. The Undertow library fails to properly validate the Host header in incoming HTTP requests.As a result, requests containing malformed or malicious Host headers are processed without...