This article examines the security implications of the Model Context Protocol (MCP) sampling feature in the context of a widely used coding copilot application.
MCP is a standard for connecting large language model (LLM) applications to external data sources and tools. We show that, without proper safeguards, malicious MCP servers can exploit the sampling feature for a range of attacks. We demonstrate these risks in practice through three proof-of-concept (PoC) examples conducted within the coding copilot, and discuss strategies for effective prevention.
Read more…
Source: Palo Alto Unit 42
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- New HiddenWasp malware found targeting Linux systems
May 29, 2019
Security researchers have found a new strain of Linux malware that appears to have been created by Chinese hackers and has been used as a means to remotely control infected systems. Named HiddenWasp, this malware is composed of a user-mode rootkit, a trojan, and an initial deployment script. The malware has a similar structure to another recently-discovered ...
- Gatekeeper Bug in MacOS Mojave Allows Malware to Execute
May 28, 2019
Researcher discloses vulnerability in macOS Gatekeeper security feature that allows the execution of malicious code on current version of the OS. Researcher Filippo Cavallarin disclosed a bug in the macOS security feature Gatekeeper that allows malicious code execution on systems running the most recent version of Mojave (10.14.0). MacOS Gatekeeper is an Apple security feature that enforces ...
- One Million Devices Open to Wormable Microsoft BlueKeep Flaw
May 28, 2019
Researchers have discovered one million devices that are vulnerable to a “wormable” Microsoft flaw, which could open the door to a WannaCry-like cyberattack. One million devices are still vulnerable to BlueKeep, a critical Microsoft bug with “wormable” capabilities, almost two weeks after a patch was released. The flaw (CVE-2019-0708) was fixed during Microsoft’s May Patch Tuesday Security Bulletin earlier this ...
- Intense scanning activity detected for BlueKeep RDP flaw
May 26, 2019
Threat actors have started scanning the internet for Windows systems that are vulnerable to the BlueKeep (CVE-2019-0708) vulnerability. This vulnerability impacts the Remote Desktop Protocol (RDP) service included in older versions of the Windows OS, such as XP, 7, Server 2003, and Server 2008. Microsoft released fixes for this vulnerability on May 14, as part of the ...
- Hackers are scanning for MySQL servers to deploy GandCrab ransomware
May 24, 2019
At least one Chinese hacking crew is currently scanning the internet for Windows servers that are running MySQL databases so they can infect these systems with the GandCrab ransomware. These attacks are somewhat unique, as cyber-security firms have not seen any threat actor until now that has attacked MySQL servers running on Windows systems to infect ...
- UK says it warned 16 NATO allies of Russian hacking activities
May 23, 2019
The UK has shared information on Russian hacking attacks with 16 NATO allies over the last 18 months, a British government official said today. “I can disclose that in the last 18 months, the National Cyber Security Centre has shared information and assessments with 16 NATO Allies – and even more nations outside the Alliance – ...