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 Very Powerful Clipboard: Analysis of a Samsung in-the-wild exploit chain
November 4, 2022
Note: The three vulnerabilities discussed in this blog were all fixed in Samsung’s March 2021 release. They were fixed as CVE-2021-25337, CVE-2021-25369, CVE-2021-25370. To ensure your Samsung device is up-to-date under settings you can check that your device is running SMR Mar-2021 or later. As defenders, in-the-wild exploit samples give us important insight into what attackers ...
- Robin Banks phishing service returns to steal banking accounts
November 4, 2022
The Robin Banks phishing-as-a-service (PhaaS) platform is back in action with infrastructure hosted by a Russian internet company that offers protection against distributed denial-of-service (DDoS) attacks. Robin Banks faced operational disruption in July 2022, when researchers at IronNet exposed the platform as a highly threatening phishing service targeting Citibank, Bank of America, Capital One, Wells Fargo, ...
- ACSC Annual Cyber Threat Report, July 2021 to June 2022
November 4, 2022
The Annual Cyber Threat Report is ACSC’s flagship unclassified publication. The Report provides an overview of key cyber threats impacting Australia, how the ACSC is responding to the threat environment, and crucial advice for Australian individuals and organisations to protect themselves online. Read more… Source: Australian Cyber Security Centre
- OPERA1ER APT in Africa
November 3, 2022
In 2019, Group-IB Threat Intelligence team detected a series of targeted attacks on financial organizations in Africa. Later in 2020, our professionals in collaboration with Orange, managed to piece together the seemingly disparate attacks into a single timeline and successfully attribute them to the threat actor codenamed OPERA1ER (also known as DESKTOP-GROUP, Common Raven, NXSMS). In ...
- Stopping C2 communications in human-operated ransomware through network protection
November 3, 2022
Command-and-control (C2) servers are an essential part of ransomware, commodity, and nation-state attacks. They are used to control infected devices and perform malicious activities like downloading and launching payloads, controlling botnets, or commanding post-exploitation penetration frameworks to breach an organization as part of a ransomware attack. Blocking these communications can mitigate attacks, sometimes before they’re ...
- Black Basta ransomware gang linked to the FIN7 hacking group
November 3, 2022
Security researchers at Sentinel Labs have uncovered evidence that links the Black Basta ransomware gang to the financially motivated hacking group FIN7, also known as “Carbanak.” When analyzing tools used by the ransomware gang in attacks, the researchers found signs that a developer for FIN7 has also authored the EDR (Endpoint Detection and Response) evasion tools ...