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:
- Ransomware Actor Abuses Genshin Impact Anti-Cheat Driver to Kill Antivirus
August 24, 2022
There have already been reports on code-signed rootkits like Netfilter, FiveSys, and Fire Chili. These rootkits are usually signed with stolen certificates or are falsely validated. However, when a legitimate driver is used as a rootkit, that’s a different story. Such is the case of mhyprot2.sys, a vulnerable anti-cheat driver for the popular role-playing game ...
- New ‘Donut Leaks’ extortion gang linked to recent ransomware attacks
August 23, 2022
A new data extortion group named ‘Donut Leaks’ is linked to recent cyberattacks, including those on Greek natural gas company DESFA, UK architectural firm Sheppard Robson, and multinational construction company Sando. Two victims disclosed these attacks without much information regarding who was involved. Over the weekend, DESFA confirmed they suffered a cyberattack after Ragnar Locker leaked screenshots ...
- Legitimate SaaS Platforms Being Used to Host Phishing Attacks
August 23, 2022
Instead of creating phishing pages from scratch, more and more cybercriminals are now abusing legitimate software-as-a-service (SaaS) platforms, including various website builders or form builders, to host their phishing pages. Since these URLs are hosted on legitimate domains, they can be especially difficult for many phishing detection engines to detect. Furthermore, these platforms typically require ...
- Colorado: Cyber attack hits Fremont County government
August 23, 2022
Fremont County government services are being impacted by a cyber attack that began last week. According to a Facebook post made by Fremont County Emergency Management, county officials became aware of the attack, which was impacting county government systems, on Wednesday, Aug. 17. An incident response team led by Fremont County Emergency Management and the Governor’s Office ...
- New Iranian APT data extraction tool
August 23, 2022
As part of TAG’s mission to counter serious threats to Google and our users, they’ve analyzed a range of persistent threats including APT35 and Charming Kitten, an Iranian government-backed group that regularly targets high risk users. For years, Google TAG have been countering this group’s efforts to hijack accounts, deploy malware, and their use of ...
- French hospital hit by $10M ransomware attack, sends patients elsewhere
August 23, 2022
The Center Hospitalier Sud Francilien (CHSF), a 1000-bed hospital located 28km from the center of Paris, suffered a cyberattack on Sunday, which has resulted in the medical center referring patients to other establishments and postponing appointments for surgeries. CHSF serves an area of 600,000 inhabitants, so any disruption in its operations can endanger the health, and ...