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:
- 1,900 Signal users exposed: Twilio attacker ‘explicitly’ looked for certain numbers
August 16, 2022
The security breach at Twilio earlier this month affected at least one high-value customer, Signal, and led to the exposure of the phone number and SMS registration codes for 1,900 users of the encrypted messaging service, it confirmed. However, Signal – considered one of the better secured of all the encrypted messaging apps – claims the ...
- Three vulnerabilities in HDF5 file format could lead to remote code execution
August 16, 2022
Cisco Talos recently discovered three vulnerabilities in a library that works with the HDF5 file format that could allow an attacker to execute remote code on a targeted device. These issues arise in the libhdf5 gif2h5 tool that’s normally used to convert a GIF file to the HDF5 format, commonly used to store large amounts of ...
- Two more malicious Python packages in the PyPI
August 16, 2022
On August 8, CheckPoint published a report on ten malicious Python packages in the Python Package Index (PyPI), the most popular Python repository among software developers. The malicious packages were intended to steal developers’ personal data and credentials. Following this research, Kaspersky used their internal automated system for monitoring open-source repositories and discovered two other malicious ...
- Threat in your browser: what dangers innocent-looking extensions hold for users
August 16, 2022
Whether you want to block ads, keep a to-do list or check your spelling, browser extensions allow you to do all of the above and more, improving convenience, productivity and efficiency for free, which is why they are so popular. Chrome, Safari, Mozilla — these and many other major Web browsers — have their own ...
- RTLS systems vulnerable to MiTM attacks, location manipulation
August 16, 2022
Security researchers have uncovered multiple vulnerabilities impacting UWB (ultra-wideband) RTLS (real-time locating systems), enabling threat actors to conduct man-in-the-middle attacks and manipulate tag geo-location data. RTLS technology is widely used in industrial environments, mass transit, healthcare, and smart city applications. Its primary role is to assist in safety by defining geofencing zones using tracking tags, signal ...
- Threat Actors Exploiting Multiple CVEs Against Zimbra Collaboration Suite
August 16, 2022
The Cybersecurity and Infrastructure Security Agency (CISA) and the Multi-State Information Sharing & Analysis Center (MS-ISAC) are publishing this joint Cybersecurity Advisory (CSA) in response to active exploitation of multiple Common Vulnerabilities and Exposures (CVEs) against Zimbra Collaboration Suite (ZCS), an enterprise cloud-hosted collaboration software and email platform. CVEs currently being exploited against ZCS include: CVE-2022-24682 CVE-2022-27924 CVE-2022-27925 ...