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:
- An unknown APT actor attacking high-profile entities in Europe and Asia
June 21, 2022
ToddyCat is a relatively new APT actor that we have not been able to relate to other known actors, responsible for multiple sets of attacks detected since December 2020 against high-profile entities in Europe and Asia. We still have little information about this actor, but we know that its main distinctive signs are two formerly ...
- Russian govt hackers hit Ukraine with Cobalt Strike, CredoMap malware
June 21, 2022
The Ukrainian Computer Emergency Response Team (CERT) is warning that Russian hacking groups are exploiting the Follina code execution vulnerability in new phishing campaigns to install the CredoMap malware and Cobalt Strike beacons. The APT28 hacking group is believed to be sending emails containing a malicious document name “Nuclear Terrorism A Very Real Threat.rtf.”. The threat ...
- Magecart attacks are still around. And they are becoming more stealthy
June 21, 2022
Magecart attacks are decreasing in number but are becoming more stealthy, with researchers highlighting potential server-side blindspots in tracking them. It’s not too often you hear about Magecart attacks. In the past few years, cybersecurity incidents that hit the headlines tended to involve attacks on core utilities and critical services, state-sponsored campaigns, ransomware, massive data breaches, ...
- Avos ransomware group expands with new attack arsenal
June 21, 2022
Avos is a ransomware group first identified in 2021 initially targeting Windows machines. More recently, a new ransomware variant of AvosLocker, named after the group, is also targeting Linux environments. Well-funded and financially motivated, Avos has been active since June 2021 and follows the ransomware-as-a-service (RaaS) model, an affiliate program to recruit potential partners. The announcement ...
- Microsoft 365 credentials targeted in new fake voicemail campaign
June 20, 2022
A new phishing campaign has been targeting U.S. organizations in the military, security software, manufacturing supply chain, healthcare and pharmaceutical sectors to steal Microsoft Office 365 and Outlook credentials. The operation is ongoing and the threat actor behind it uses fake voicemail notifications to lure victims into opening a malicious HTML attachment. According to researchers at cloud ...
- Android-wiping BRATA malware is evolving into a persistent threat
June 19, 2022
The threat actor behind BRATA banking trojan has evolved their tactics and improved the malware with information-stealing capabilities. Italian mobile security company Cleafy has been tracking BRATA activity and noticed in the most recent campaigns changes that lead to longer persistence on the device. “The modus operandi now fits into an Advanced Persistent Threat (APT) activity pattern,” ...