AI security research · defensive focus · bilingual
XiaozaoSec
- 37 articles
- 7 topics
- 2026-09 last update
Latest articles
-
MCP Security Series: Authentication and Authorization Flaws
Authentication and authorization flaws broadly come down to permission problems in MCP services. Here are a few scenarios:
-
MCP Security Series: Intent Flow Subversion
Intent flow subversion occurs when the model is supposed to perform task A, but we inject instructions that cause it to perform task B instead, or to tamper with the outc…
-
MCP Security Series: Command Injection
Command injection means the MCP server itself is vulnerable to command injection and similar flaws. The malicious commands it executes can come from many sources: a user…
-
MCP Security Series: Tool Poisoning
The tool poisoning risk in MCP security actually revolves around malicious MCPs. For example, when the MCP server's code is malicious, or when an MCP tool's description c…
-
AIG: A Black Hat-Selected Open-Source AI Red Teaming Platform - Complete Guide
This post walks through how to use AIG to perform Agent security scanning, MCP security scanning, Skill security scanning, LLM security scanning, and AI application compo…
-
How to Improve Cybersecurity Awareness in the AI Era
In the AI era, generative large models have dramatically lowered the barrier to fraud. Phishing emails that used to be written by hand can now be drafted by AI, and the i…
-
MCP Security Series: Excessive Permissions
This category of issue refers primarily to MCP tools having overly broad permissions and failing to follow the principle of least privilege. It shares the same underlying…
-
MCP Security Information Leakage
The first risk in MCP, when translated literally, is actually called improper token management and key leakage. For the sake of convenience, we will simply refer to it as…
-
MCP Security Series: MCP Basics
A key part of an Agent is its ability to call external tools to get things done. For example, you could ask an Agent to plan a route from A to B for tomorrow, find restau…
-
Defending Against Malicious Skills: SkillSpector Deployment and Detection
SkillSpector is an open-source Skill security scanning tool developed by NVIDIA, with the following features:
-
Skill Security Risk: Metadata Security
In the Skill Basics article, we learned that the content between --- in a SKILL.md file is Skill metadata. This metadata typically defines various descriptive information…
-
Skill Security Series — Excessive Permissions
Excessive skill permissions refer to when a skill's granted permissions are far beyond what it actually needs. For example:
Topics
- LLM Security 20 articles How prompt injection and jailbreaks work and how to defend against them, a walkthrough of the OWASP LLM Top 10, plus model supply-chain and data-poisoning risks.
- AI Agent Security 1 articles Agent permission boundaries and excessive-agency risk, the OWASP Agent Top 10, and the security of multi-agent collaboration and the A2A protocol.
- MCP Security 7 articles MCP protocol fundamentals, tool poisoning, command injection, excessive permissions and information leakage — the OWASP MCP Top 10 explained and defended.
- AI Skills Security 7 articles SKILL fundamentals, metadata risk and over-permissioning, and how to identify and detect malicious skills and supply-chain attacks.
- AI-Assisted Cybersecurity 2 articles Putting AI to work on the defence side: assisted vulnerability discovery, code audit and security operations, plus security awareness in the AI era.
- Robotics Security In progress Security for humanoid robots and embodied AI: communication links, sensor spoofing, actuation permissions and the blast radius in the physical world.
- AI Data Security In progress Privacy and integrity of AI data: leakage and tampering risks across training corpora, vector stores, retrieval augmentation and the inference path.