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:
- Cybersecurity Researchers Spotted First-Ever UEFI Rootkit in the Wild
September 27, 2018
Cybersecurity researchers at ESET have unveiled what they claim to be the first-ever UEFI rootkit being used in the wild, allowing hackers to implant persistent malware on the targeted computers that could survive a complete hard-drive wipe. Dubbed LoJax, the UEFI rootkit is part of a malware campaign conducted by the infamous Sednit group, also known as ...
- Cobalt threat group serves up SpicyOmelette in fresh bank attacks
September 27, 2018
Advanced persistent threat group (APT) the Cobalt Gang, also known as Gold Kingswood, is spreading SpicyOmelette malware in campaigns targeting financial institutions worldwide. In a world where cyberattacks against businesses and consumers alike are spreading and evolving in nature and sophistication, it is often financial institutions which bear the brunt. Banking customers hoodwinked by fraudulent schemes or ...
- VPNFilter’s Arsenal Expands With Newly Discovered Modules
September 26, 2018
Seven new modules discovered in VPNFilter further fill in the blanks about how the malware operates and reveals a wider breath of capabilities. Researchers have discovered new modules in VPNFilter – the malware behind the widespread campaign in May that infected 75 router brands – revealing that its capabilities are much more widespread and sophisticated than previously thought. After ...
- New Linux Kernel Bug Affects Red Hat, CentOS, and Debian Distributions
September 26, 2018
Security researchers have published the details and proof-of-concept (PoC) exploits of an integer overflow vulnerability in the Linux kernel that could allow an unprivileged user to gain superuser access to the targeted system. The vulnerability, discovered by cloud-based security and compliance solutions provider Qualys, which has been dubbed “Mutagen Astronomy,” affects the kernel versions released between ...
- Over 80 Cisco Products Affected by FragmentSmack DoS Bug
September 25, 2018
Cisco is currently looking into its product line to determine which products and services use Linux kernel 3.9 or above, which is vulnerable to the FragmentSmack denial-of-service (DoS) bug. The networking hardware manufacturer already assembled a list of more than 80 products that are affected by the vulnerability. Many of them expect a fix by February ...
- macOS Mojave Privacy Bypass Flaw Allows Access to Protected Files
September 24, 2018
A security researcher shows on Mojave’s release day that Apple’s latest privacy protection implementations in macOS are not sufficiently strong. In a minute-long clip, Patrick Wardle shows that the security in the dark-themed macOS can be bypassed to reach sensitive user data, such as the information in the address book. Talking to BleepingComputer, Wardle says that he ...