Shiny tools, shallow checks: how the AI hype opens the door to malicious MCP servers


In this article, Kaspersky researchers explore how the Model Context Protocol (MCP) — the new “plug-in bus” for AI assistants — can be weaponized as a supply chain foothold.

The researchers start with a primer on MCP, map out protocol-level and supply chain attack paths, then walk through a hands-on proof of concept: a seemingly legitimate MCP server that harvests sensitive data every time a developer runs a tool. Then they break down the source code to reveal the server’s true intent and provide a set of mitigations for defenders to spot and stop similar threats.

Read more…
Source: Kaspersky


Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox


Related:

  • Apple removes feature that allowed its apps to bypass macOS firewalls and VPNs

    January 14, 2021

    Apple has removed a controversial feature from the macOS operating system that allowed 53 of Apple’s own apps to bypass third-party firewalls, security tools, and VPN apps installed by users for their protection. Known as the ContentFilterExclusionList, the list was included in macOS 11, also known as Big Sur. The exclusion list included some of Apple’s biggest ...

  • CISCO says it won’t patch 74 security bugs in older RV routers that reached EOL

    January 14, 2021

    Networking equipment vendor Cisco said yesterday it was not going to release firmware updates to fix 74 vulnerabilities that had been reported in its line of RV routers, which had reached end-of-life (EOL). Affected devices include Cisco Small Business RV110W, RV130, RV130W, and RV215W systems, which can be used as both routers, firewalls, and VPNs. All four ...

  • The Top Worry In Cloud Security for 2021

    January 13, 2021

    The cloud is an environment full of potential. It provides easy access to technologies that simple weren’t available a decade ago. You can now launch the equivalent of an entire data center with a single command. Scaling to meet the demands of millions of customers can be entirely automated. Advanced machine learning analysis is as simple ...

  • Sophisticated Hacks Against Android, Windows Reveal Zero-Day Trove

    January 13, 2021

    Google researchers have detailed a major hacking campaign that was detected in early 2020, which mounted a series of sophisticated attacks, some using zero-day flaws, against Windows and Android platforms. Working together, researchers from Google Project Zero and the Google Threat Analysis Group (TAG) uncovered the attacks, which were “performed by a highly sophisticated actor,” Ryan ...

  • Hackers leak stolen Pfizer COVID-19 vaccine data online

    January 12, 2021

    The European Medicines Agency (EMA) today revealed that some of the Pfizer/BioNTech COVID-19 vaccine data stolen from its servers in December was leaked online. EMA is a decentralized agency responsible for reviewing and approving COVID-19 vaccines, as well as for evaluating, monitoring, and supervising any new medicines introduced to the EU. “The ongoing investigation of the cyberattack ...

  • Critical Microsoft Defender Bug Actively Exploited; Patch Tuesday Offers 83 Fixes

    January 12, 2021

    Microsoft addressed 10 critical bugs, one under active exploit and another publicly known, in its January Patch Tuesday roundup of fixes. In total it patched 83 vulnerabilities. The most serious bug is a flaw in Microsoft’s Defender anti-malware software that allows remote attackers to infect targeted systems with executable code. Security experts are warning that Windows ...