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:
- A Post-exploitation Look at Coinminers Abusing WebLogic Vulnerabilities
September 14, 2022
Trend Micro researchers have recently observed malicious actors exploiting both recently disclosed and older Oracle WebLogic Server vulnerabilities to deliver cryptocurrency-mining malware. Oracle WebLogic Server is typically used for developing and deploying high-traffic enterprise applications on cloud environments and engineered and conventional systems. One of the older vulnerabilities that is still being actively exploited by malicious ...
- FBI: Cyber Criminals Targeting Healthcare Payment Processors, Costing Victims Millions in Losses
September 14, 2022
The FBI has received multiple reports of cyber criminals increasingly targeting healthcare payment processors to redirect victim payments. In each of these reports, unknown cyber criminals used employees’ publicly-available Personally Identifiable Information (PII) and social engineering techniques to impersonate victims and obtain access to files, healthcare portals, payment information, and websites. In one case, the attacker ...
- New PsExec spinoff lets hackers bypass network security defenses
September 13, 2022
Security researchers have developed an implementation of the Sysinternals PsExec utility that allows moving laterally in a network using a single, less monitored port, Windows TCP port 135. PsExec is designed to help administrators execute processes remotely on machines in the network without the need to install a client. Threat actors have also adopted the tool and ...
- New Wave of Espionage Activity Targets Asian Governments
September 13, 2022
A distinct group of espionage attackers who were formerly associated with the ShadowPad remote access Trojan (RAT) has adopted a new, diverse toolset to mount an ongoing campaign against a range of government and state-owned organizations in a number of Asian countries. The attacks, which have been underway since at least early 2021, appear to ...
- Cisco confirms Yanluowang ransomware leaked stolen company data
September 12, 2022
Cisco has confirmed that the data leaked yesterday by the Yanluowang ransomware gang was stolen from the company network during a cyberattack in May. However, the company says in an update that the leak does not change the initial assessment that the incident has no impact on the business: Read more… Source: Bleeping Computer
- Shape-shifting cryptominer savaging Linux endpoints and IoT
September 10, 2022
AT&T cybersecurity researchers have discovered a sneaky piece of malware targeting Linux endpoints and IoT devices in the hopes of gaining persistent access and turning victims into crypto-mining drones. The malware was dubbed “Shikitega” for its extensive use of the popular Shikata Ga Nai polymorphic encoder, which allows the malware to “mutate” its code to avoid ...