Ghostcommit is a proof of concept that shows how AI assistants used to review software code can be tricked by hidden instructions embedded in images.
The academic ASSET Research Group showed that an attacker can place instructions inside an image file, point to it in an AGENTS.md file, and get an AI coding agent to follow those instructions during a later task.
A pull request is basically a formal “please review and add my changes” request that a developer sends before changes are added to the main version of a software project. Human reviewers and, increasingly, AI coding tools may review the changes before they’re accepted.
Read more…
Source: MalwareBytes Labs
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- Operational Technology Widens Supply Chain Attack Surfaces
March 8, 2019
Today’s supply chain has evolved, with operational technology (OT) used in factories increasingly becoming connected and converging with IT systems — introducing new attack vectors. This new reality is vital for companies to understand in the context of risk, according to Dawn Cappelli, vice president of global security and CISO at Rockwell Automation and Edna Conway, ...
- New SLUB Backdoor Uses GitHub, Communicates via Slack
March 7, 2019
We recently came across a previously unknown malware that piqued our interest in multiple ways. For starters, we discovered it being spread via watering hole attacks, a technique that involves an attacker compromising a website before adding code to it so visitors are redirected to the infecting code. In this case, each visitor is redirected only ...
- Data-Wiping Cyberattacks Plague Financial Firms
March 6, 2019
Over a quarter of surveyed financial institutions reported that they were targeted by destructive cyberattacks over the past year, bent on completely destroying data. That’s according to a new Carbon Black report unveiled at RSA this year. The report, “Modern Bank Heists: The Bank Robbery Shifts to Cyberspace,” outlines the top attacks that financial firms are facing ...
- IoT Devices Under Constant Attack
March 4, 2019
‘Secure your IoT devices’ is the message from security specialist Cyxtera Technologies, after research found that IoT devices are now under constant attack. The research was conducted jointly by Cyxtera threat researcher Martin Ochoa and researchers from the Singapore University of Technology and Design. They detected more than 150 million connection attempts to 4,642 distinct IP addresses ...
- How the Dark Web Data Bazaar Fuels Enterprise Attacks
March 3, 2019
It seems every aspect of our lives is available to be found somewhere on the internet. And the information available isn’t simply embarrassing browsing histories but ranges from our medical histories to the logon credentials we use to access many of our online services. This is certainly a privacy concern, but it’s also increasingly an enterprise ...
- New exploit lets attackers take control of Windows IoT Core devices
March 2, 2019
Speaking at a conference today, a security researcher has revealed a new exploit impacting the Windows IoT Core operating system that gives threat actors full control over vulnerable devices. The vulnerability, discovered by Dor Azouri, a security researcher for SafeBreach, impacts the Sirep/WPCon communications protocol included with Windows IoT operating system. Azouri said the vulnerability only impacts Windows ...