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:
- Old Tools for New Money: URL Spreading Shellbot and XMRig Using 17-year old XHide
July 19, 2019
One of our honeypots detected a threat that propagates by scanning for open ports and brute forcing weak credentials, installing a Monero cryptocurrency miner and a Perl-based IRC backdoor as the final payload. The miner process is hidden using XHide Process Faker, a 17-year old open source tool used to fake the name of a ...
- Iran-Linked APT34 Invites Victims to LinkedIn for Fresh Malware Infections
July 19, 2019
A recent phishing campaign by Iran-linked threat actor APT34 made use of a savvy approach: Asking victims to join their social network. According to FireEye, the adversaries masqueraded as a Cambridge University lecturer, including setting up a LinkedIn page, in order to gain victims’ trust. From there the attackers asked their “friends” to open malicious documents. APT34, ...
- Spam Campaign Targets Colombian Entities with Custom-made ‘Proyecto RAT,’ Uses Email Service YOPmail for C&C
July 18, 2019
We observed a recent campaign that primarily targets financial institutions and governmental organizations in the South American region, particularly in Colombia. This blog post covers the activities we observed, the remote access tools (RATs) used, the campaign’s techniques and procedures, and its indicators of compromise (IoCs). Our findings indicate that the campaign appears to be ...
- Mirai Botnet Sees Big 2019 Growth, Shifts Focus to Enterprises
July 18, 2019
The infamous Mirai internet of things botnet is spiking in growth while changing up its tactics, techniques and procedures so far in 2019, to target more and more enterprise-level hardware, It’s a state of affairs that presents a greater concern than ever before given the ongong migration to the cloud era, researchers said. According to researchers ...
- StrongPity APT Returns with Retooled Spyware
July 17, 2019
The APT group behind the sophisticated malware known as StrongPity (a.k.a. Promethium) has mounted a fresh spyware campaign that is still ongoing as of July 2019. The group has retooled with new malware to control compromised machines, according to researchers. “The new malware samples have been unreported and generally appear to ...
- Why Cities Are a Low-Hanging Fruit For Ransomware
July 15, 2019
Ransomware attacks against local governments and cities are repeatedly making headlines, with crippling results on city operations and budgets. Last month, the Florida city of Riviera Beach paid hackers $600,000 after being hit by a ransomware attack that downed its computer systems for three weeks. In 2018, several Atlanta city systems were crippled after a ransomware attack extorted ...