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:
- FBI: Hackers stole source code from US government agencies and private companies
November 7, 2020
The Federal Bureau of Investigation has sent out a security alert warning that threat actors are abusing misconfigured SonarQube applications to access and steal source code repositories from US government agencies and private businesses. Intrusions have taken place since at least April 2020, the FBI said in an alert sent out last month and made public ...
- When Threat Actors Fly Under the Radar: Vatet, PyXie and Defray777
November 6, 2020
As security practitioners, Palo Alto Unit 42 researchers spend a lot of time focusing on the threat actors and malware families that leverage the most impactful exploits or affect the highest number of victims. But what happens when a threat actor goes “low and slow” to fly under the radar? One could argue that, in ...
- RansomEXX Trojan attacks Linux systems
November 6, 2020
Kaspersky researchers have recently discovered a new file-encrypting Trojan built as an ELF executable and intended to encrypt data on machines controlled by Linux-based operating systems. After the initial analysis we noticed similarities in the code of the Trojan, the text of the ransom notes and the general approach to extortion, which suggested that we had ...
- Apple Patches Bugs Tied to Previously Identified Zero-Days
November 6, 2020
Apple has patched three previously identified zero-day vulnerabilities in its iPhone, iPod and iPad devices potentially related to a spate of related flaws recently discovered by the Google Project Zero team that also affect Google Chrome and Windows. Apple this week released iOS 14.2 and iPadOS 14.2, which patch a total of 24 vulnerabilities—including the three ...
- Operation North Star: Behind The Scenes
November 5, 2020
It is rare to be provided an inside view on how major cyber espionage campaigns are conducted within the digital realm. The only transparency afforded is a limited view of victims, a malware sample, and perhaps the IP addresses of historical command and control (C2) infrastructure. The Operation North Star campaign we detailed earlier this year ...
- Italian beverage vendor Campari knocked offline after ransomware attack
November 5, 2020
Campari Group, the famed Italian beverage vendor behind brands like Campari, Cinzano, and Appleton, has been hit by a ransomware attack and has taken down a large part of its IT network. The attack took place last Sunday, on November 1, and has been linked to the RagnarLocker ransomware gang, according to a copy of the ...