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:
- Google says Apple employee found a zero-day but did not report it
July 20, 2023
Google fixed a zero-day in Chrome that was found by an Apple employee, according to comments in the official bug report. While the bug itself is not newsworthy, the circumstances of how this bug was found and reported to Google are, to say the least, peculiar. According to a Google employee, the bug was originally found ...
- KillNet Showcases New Capabilities While Repeating Older Tactics
July 20, 2023
In early 2022, Mandiant predicted that Russian cyber threat activity associated with the invasion of Ukraine would affect government and private sector targets in third-party countries, particularly neighboring countries, North Atlantic Treaty Organization (NATO) allies, and other nations voicing support for Ukraine. Russian government-linked actors have historically employed false hacktivist facades as a means of ...
- A nasty Google Cloud bug could let hackers use it to launch attacks
July 20, 2023
Cybersecurity researchers from Orca Security have uncovered a new bug in the Google Cloud Build service which could allow threat actors to gain almost full access to Google Artifact Registry code repositories. The repercussions of the flaw, the researchers are saying in their report, are quite dire. The researchers named the vulnerability Bad.Build, saying it allows ...
- CISA Releases Cybersecurity Advisory on Threat Actors Exploiting Citrix CVE-2023-3519
July 20, 2023
The Cybersecurity and Infrastructure Security Agency (CISA) released a Cybersecurity Advisory (CSA), Threat Actors Exploiting Citrix CVE-2023-3519 to Implant Webshells, to warn organizations about threat actors exploiting CVE-2023-3519, an unauthenticated remote code execution (RCE) vulnerability affecting NetScaler (formerly Citrix) Application Delivery Controller (ADC) and NetScaler Gateway. In June 2023, threat actors exploited this vulnerability as a ...
- North Korea-backed hackers breached JumpCloud to target cryptocurrency clients
July 20, 2023
North Korean state-backed hackers breached U.S. enterprise software company JumpCloud to target its cryptocurrency clients, security researchers said on Thursday. JumpCloud, a directory platform that allows enterprises to authenticate, authorize and manage users and devices, said this week that a nation-state actor was behind a June breach of its systems that forced the company to reset ...
- Comprehensive analysis of initial attack samples exploiting CVE-2023-23397 vulnerability
July 19, 2023
On March 14, 2023, Microsoft published a blogpost describing an Outlook Client Elevation of Privilege Vulnerability (CVSS: 9.8 CRITICAL). The publication generated a lot of activity among white, grey and black hat researchers, as well as lots of publications and tweets about the vulnerability and its exploitation. In this post Kaspersky researchers highlight the key points ...