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:
- Midnight Blizzard conducts targeted social engineering over Microsoft Teams
August 2, 2023
Microsoft Threat Intelligence has identified highly targeted social engineering attacks using credential theft phishing lures sent as Microsoft Teams chats by the threat actor that Microsoft tracks as Midnight Blizzard (previously tracked as NOBELIUM). This latest attack, combined with past activity, further demonstrates Midnight Blizzard’s ongoing execution of their objectives using both new and common ...
- Sha zhu pan scam uses AI chat tool to target iPhone and Android users
August 2, 2023
Over the past two years, we have been tracking a variety of scams targeting mobile device users, generally referred to as “shā zhū pán” (杀猪盘, which translates as “butcher plate”) or “Pig Butchering.” This includes a category we labelled as “CryptoRom” when we initially investigated it in 2020, because of its two distinguishing characteristics—a focus on ...
- CISA and International Partner NCSC-NO Release Joint Cybersecurity Advisory on Threat Actors Exploiting Ivanti EPMM Vulnerabilities
August 1, 2023
The Cybersecurity and Infrastructure Security Agency (CISA) and the Norwegian National Cyber Security Centre (NCSC-NO) have released a joint Cybersecurity Advisory (CSA), Threat Actors Exploiting Ivanti EPMM Vulnerabilities, in response to the active exploitation of CVE-2023-35078 and CVE-2023-35081 affecting Ivanti Endpoint Manager Mobile (EPMM) (formerly known as MobileIron Core). Threat actors can chain these vulnerabilities to ...
- NodeStealer 2.0 – The Python Version: Stealing Facebook Business Accounts
August 1, 2023
Unit 42 researchers have recently discovered a previously unreported phishing campaign that distributed an infostealer equipped to fully take over Facebook business accounts. Facebook business accounts were targeted with a phishing lure offering tools such as spreadsheet templates for business. This is part of a growing trend of threat actors targeting Facebook business accounts – for ...
- Thailand feels the force of cyber-attacks
August 1, 2023
The average number of cyber-attacks on organisations in Thailand was almost double the average rate globally and slightly higher than the average within Southeast Asia over the past six months, according to Check Point Research. Thai organisations were attacked 2,388 times per week on average during the last six months, compared with 2,375 attacks per week ...
- Google AMP – The Newest of Evasive Phishing Tactic
August 1, 2023
A new phishing tactic utilizing Google Accelerated Mobile Pages (AMP) has hit the threat landscape and proven to be very successful at reaching intended targets. Google AMP is an open-source HTML framework used to build websites that are optimized for both browser and mobile use. The websites that Cofense researches observed in these campaigns are hosted ...