New Prompt Injection Attack Vectors Through MCP Sampling


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:

  • Hackers trying to corrupt AI, raising level of ransomware threat

    October 17, 2023

    Cyber criminals are actively trying to corrupt generative artificial intelligence (AI), which may then put the ability to create ransomware in the hands of individuals. The looming threat is what keeps Mr Willis Lim, the director of the National Cyber Threat Analysis Centre at the Cyber Security Agency of Singapore (CSA), up at night. Generative ...

  • Over 70pc of Irish businesses suffered cyber attack in the past 12 months – report

    October 17, 2023

    Over 70pc of Irish businesses have suffered a cyber attack over the past year, a new report from insurer Hiscox has revealed. Around 71pc of the 200 Irish businesses surveyed experienced a cyber attack in the past 12 months, up 22pc from the same period last year. Ireland also had the highest median average number ...

  • Kaspersky uncovers APT campaign targeting APAC government entities

    October 17, 2023

    Kaspersky researchers have discovered a persistent campaign compromising a specific type of secure USB drive used to provide encryption for safe data storage. Dubbed “TetrisPhantom,” this espionage effort targets government entities in the Asia-Pacific region (APAC), and shows no discernible overlap with any known threat actor. These and other findings are detailed in Kaspersky’s new ...

  • Urgent global response needed for “insidious” cybercrime – Interpol

    October 16, 2023

    SINGAPORE – New types of cybercrime are emerging all the time. Manipulative and well-organized cybercriminals are exploiting digital technologies to tailor their attacks and target weaknesses in online systems, networks and infrastructures. The complex and borderless nature of cybercrime is compounded by the involvement of transnational organized crime groups, underlining the need to mount an ...

  • Thailand: House of Representatives’ Website Hacked, Cyber Attack Investigation Underway

    October 16, 2023

    The House of Representatives’ website fell victim to a cyber attack on Sunday, October 15, 2023. The hackers, who go by the name 3MUSKETEERZ, managed to breach the website’s security and display a picture of a troll in the photo journal section. Additionally, the perpetrators altered the press releases and committee schedules featured on the site. ...

  • Understanding DNS Tunneling Traffic in the Wild

    October 13, 2023

    Palo Alto Unit 42 researchers present a study on why and how domain name system (DNS) tunneling techniques are used in the wild. Motivated by their findings, they present a system to automatically attribute tunneling domains to tools and campaigns. Attackers adopt DNS tunneling techniques to bypass security policies in enterprise networks because most enterprises ...