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:
- Storm-2372 conducts device code phishing campaign
February 13, 2025
Microsoft discovered cyberattacks being launched by a group they call Storm-2372, who they assess with medium confidence aligns with Russia’s interests and tradecraft. The attacks appear to have been ongoing since August 2024 and have targeted governments, NGOs, and a wide range of industries in multiple regions. The attacks use a specific phishing technique called “device ...
- Spyware maker caught distributing malicious Android apps for years
February 13, 2025
Italian spyware maker SIO, known to sell its products to government customers, is behind a series of malicious Android apps that masquerade as WhatsApp and other popular apps but steal private data from a target’s device, TechCrunch has exclusively learned. Late last year, a security researcher shared three Android apps with TechCrunch, claiming they were likely ...
- Upper Michigan: Cyber attack hits Sault Tribe offices
February 13, 2025
A ransomware attack that shut down gaming at all five Kewadin Casino locations also impacted other offices at an eastern Upper Peninsula tribe. The tribe made the announcement Monday and said it could be a week or more before regular operations can resume. “On Sunday morning, the Sault Ste. Marie Tribe of Chippewa Indians suffered a ...
- Ivanti Releases February 2025 Security Updates
February 12, 2025
Ivanti has released three security advisories in the February Security Update, which addresses vulnerabilities in Ivanti products. In the first advisory, two vulnerabilities were identified in Ivanti Cloud Services Application (CSA). The Ivanti CSA is an Internet appliance that provides secure communication and functionality over the Internet. It falls under the primary product of Ivanti Endpoint ...
- SonicOS SSL VPN Authentication Bypass Vulnerability (CVE-2024-53704)
February 12, 2025
A proof-of-concept (PoC) exploit has been published by security researchers for an authentication bypass vulnerability in the SonicOS SSL VPN component. SonicWall appliances provide virtual private network (VPN) and ‘next-gen’ firewall capabilities. SonicWall formally disclosed and released security updates addressing CVE-2024-53704 on 07 January 2025. Successful exploitation of CVE-2024-53704 could allow a remote, unauthenticated attacker to ...
- The BadPilot campaign: Seashell Blizzard subgroup conducts multiyear global access operation
February 12, 2025
Microsoft is publishing for the first time their research into a subgroup within the Russian state actor Seashell Blizzard and its multiyear initial access operation, tracked by Microsoft Threat Intelligence as the “BadPilot campaign”. This subgroup has conducted globally diverse compromises of Internet-facing infrastructure to enable Seashell Blizzard to persist on high-value targets and support tailored ...