Model Context Protocol (MCP) Archives

FakeGit Campaign Uses 7,600 GitHub Repositories to Spread SmartLoader Malware

20Jul

By RBNo Comments

News

FakeGit Campaign Uses 7,600 GitHub Repositories to Spread SmartLoader Malware

As part of an ongoing campaign known as FakeGit, cybersecurity researchers have found over 7,600 malicious GitHub repositories, of which more than 800 pose as Model Context Protocol (MCP) servers or artificial intelligence (AI) talents to distribute a malware family known as SmartLoader. According to a report shared with The Hacker News by Oleg Zaytsev, principal security researcher at Island, FakeGit employs malicious ZIP files, lookalike developer profiles, replicated projects, and persuasive READMEs to spread SmartLoader malware. The ultimate objective of these assaults is to use the access provided by SmartLoader to create persistence and propagate secondary payloads, like StealC, an information stealer that can extract a variety of data from infected systems. It's important ...

Critical Nginx UI auth bypass flaw now actively exploited in the wild

A serious flaw in the Nginx UI that supports Model Context Protocol (MCP) is already being used in the field to fully take over a server without requiring authentication. The vulnerability, known as CVE-2026-33032, is brought about by nginx-ui leaving the '/mcp_message' endpoint unprotected, which enables remote attackers to use privileged MCP activities without the need for credentials. One unauthenticated request can alter server behavior and essentially take over the web server because those activities entail writing and reloading nginx configuration files. According to NIST's description of the vulnerability in the National Vulnerability Database (NVD), "any network attacker can invoke all MCP tools without authentication, including restarting nginx, creating/modifying/deleti...

Researchers Demonstrate How MCP Prompt Injection Can Be Used for Both Attack and Defense

According to a recent report from Tenable, new research has discovered how methods that make the Model Context Protocol (MCP) vulnerable to prompt injection attacks could be used to create security tooling or detect malicious tools, as the field of artificial intelligence (AI) continues to advance at a rapid pace. In order to improve the precision, applicability, and usefulness of AI applications, Anthropic introduced MCP in November 2024. It is a framework for connecting Large Language Models (LLMs) with external data sources and services and using model-controlled tools to communicate with those systems. Its client-server design enables hosts with MCP clients, like Cursor or Claude Desktop, to connect to various MCP servers, each of which makes available particular tools and funct...