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:
- BPFdoor: Stealthy Linux malware bypasses firewalls for remote access
May 12, 2022
A recently discovered backdoor malware called BPFdoor has been stealthily targeting Linux and Solaris systems without being noticed for more than five years. BPFdoor is a Linux/Unix backdoor that allows threat actors to remotely connect to a Linux shell to gain complete access to a compromised device. The malware does not need to open ports, it can’t ...
- APT34 hackers exposed in a highly targeted espionage campaign
May 12, 2022
Threat analysts have spotted a novel attack attributed to the Iranian hacking group known as APT34 group or Oilrig, who targeted a Jordanian diplomat with custom-crafted tools. The attack involved advanced anti-detection and anti-analysis techniques and had some characteristics that indicate lengthy and careful preparation. Security researchers at Fortinet have gathered evidence and artifacts from the attack ...
- New IceApple exploit toolset deployed on Microsoft Exchange servers
May 11, 2022
Security researchers have found a new post-exploitation framework that they dubbed IceApple, deployed mainly on Microsoft Exchange servers across a wide geography. IceApple is described as being “highly sophisticated,” its developer prioritizing keeping a low profile for long-term objectives in targeted attacks. The framework was discovered by the Falcon OverWatch team, CrowdStrike’s proactive threat hunting division, in ...
- Microsoft closes Windows LSA hole under active attack
May 11, 2022
Microsoft patched 74 security flaws in its May Patch Tuesday batch of updates. That’s seven critical bugs, 66 deemed important, and one ranked low severity. At least one of the vulnerabilities disclosed is under active attack with public exploit code, according to Redmond, while two others are listed as having public exploit code. After April’s astonishing 100-plus ...
- CISA: Protecting Against Cyber Threats to Managed Service Providers and their Customers
May 11, 2022
The cybersecurity authorities of the United Kingdom (NCSC-UK), Australia (ACSC), Canada (CCCS), New Zealand (NCSC-NZ), and the United States (CISA), (NSA), (FBI) are aware of recent reports that observe an increase in malicious cyber activity targeting managed service providers (MSPs) and expect this trend to continue. This joint Cybersecurity Advisory (CSA) provides actions MSPs and ...
- Team of experts help Rutube to recover from the May 9 cyberattack
May 11, 2022
Rutube involved several expert teams, including a team of specialists from Positive Technologies security center, to deal with the aftermath of the May 9 cyberattack, the website said in its Telegram channel. “In order to investigate the attack and deal with its aftermath, several expert teams were involved, including a team of specialists from the Positive ...