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:
- Hackers target Japanese politicians with new MirrorStealer malware
December 15, 2022
A hacking group tracked as MirrorFace has been targeting Japanese politicians for weeks before the House of Councilors election in July 2022, using a previously undocumented credentials stealer named ‘MirrorStealer.’ The campaign was discovered by ESET, whose analysts report they could piece together evidence thanks to operational mistakes made by the hackers that left traces behind. The ...
- Reassessing cyberwarfare. Lessons learned in 2022
December 14, 2022
At this point, it has become cliché to say that nothing in 2022 turned out the way we expected. We left the COVID-19 crisis behind hoping for a long-awaited return to normality and were immediately plunged into the chaos and uncertainty of a twentieth-century-style military conflict that posed serious risks of spreading over the continent. ...
- Cyber Signals: Risks to critical infrastructure on the rise
December 14, 2022
Today, the third edition of Cyber Signals was released spotlighting security trends and insights gathered from Microsoft’s 43 trillion daily security signals and 8,500 security experts. In this edition, we share new insights on wider risks that converging IT, Internet of Things (IoT), and operational technology (OT) systems pose to critical infrastructure. Cyber Signals presents ...
- Attackers use SVG files to smuggle QBot malware onto Windows systems
December 14, 2022
QBot malware phishing campaigns have adopted a new distribution method using SVG files to perform HTML smuggling that locally creates a malicious installer for Windows. This attack is made through embedded SVG files containing JavaScript that reassemble a Base64 encoded QBot malware installer that is automatically downloaded through the target’s browser. QBot is a Windows malware arriving ...
- Probing Weaponized Chat Applications Abused in Supply-Chain Attacks
December 14, 2022
In late September 2022, threat researchers uncovered a supply-chain attack carried out by malicious actors using a trojanized installer of Comm100, a chat-based customer engagement application. Trend Micro investigation of the incident revealed that the breadth and depth of the campaign’s impact were greater than what the researchers had initially thought; Trend Micro researchers also ...
- Hackers exploit critical Citrix ADC and Gateway zero day, patch now
December 13, 2022
Citrix strongly urges admins to apply security updates for an ‘Critical’ zero-day vulnerability (CVE-2022-27518) in Citrix ADC and Gateway that is actively exploited by state-sponsored hackers to gain access to corporate networks. This new vulnerability allows an unauthenticated attacker to execute commands remotely on vulnerable devices and take control over them. Citrix is warning admins to install ...