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:
- Billbug: State-sponsored Actor Targets Cert Authority, Government Agencies in Multiple Asian Countries
November 15, 2022
State-sponsored actors compromised a digital certificate authority in an Asian country during a campaign in which multiple government agencies were also targeted. Symantec, by Broadcom Software, was able to link this activity to a group we track as Billbug due to the use in this campaign of tools previously attributed to this group. Billbug (aka Lotus ...
- Shocker: EV charging infrastructure is seriously insecure
November 15, 2022
If you’ve noticed car charging stations showing up in your area, congratulations! You’re part of a growing network of systems so poorly secured they could one day be used to destabilize entire electrical grids, and which contain enough security issues to be problematic today. That’s what scientists at Sandia National Laboratory in Albuquerque, New Mexico have ...
- DTrack activity targeting Europe and Latin America
November 15, 2022
DTrack is a backdoor used by the Lazarus group. Initially discovered in 2019, the backdoor remains in use three years later. It is used by the Lazarus group against a wide variety of targets. For example, Kaspersky researchers seen it being used in financial environments where ATMs were breached, in attacks on a nuclear power ...
- Android malware: A million people downloaded these malicious apps before they were finally removed from Google Play
November 15, 2022
Google has removed a series of apps downloaded by over a million Android users from the Google Play Store that infected smartphones with malware and bombarded devices with malicious pop-up ads. The malware has been detailed by cybersecurity researchers at Malwarebytes. The apps were still available to download for a number of days after the research ...
- Russia-based Pushwoosh tricks US Army and others into running its code – for a while
November 15, 2022
US government agencies including the Army and Centers for Disease Control and Prevention pulled apps running Pushwoosh code after learning the software company – which presents itself as American – is actually Russian, according to Reuters. Pushwoosh is a software company that provides code and data analysis for developers so they can automate custom push notifications ...
- CISA Has Added One Known Exploited Vulnerability to Catalog
November 14, 2022
CISA has added one new vulnerability to its Known Exploited Vulnerabilities Catalog, based on evidence of active exploitation. This type of vulnerability is a frequent attack vector for malicious cyber actors and pose significant risk to the federal enterprise. Note: To view the newly added vulnerabilities in the catalog, click on the arrow in the “Date ...