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:
- FIN7.5: the infamous cybercrime rig “FIN7” continues its activities
May 8, 2019
On August 1, 2018, the US Department of Justice announced that it had arrested several individuals suspected of having ties to the FIN7 cybercrime rig. FIN7 operations are linked to numerous intrusion attempts having targeted hundreds of companies since at least as early as 2015. Interestingly, this threat actor created fake companies in order to ...
- Surge of MegaCortex ransomware attacks detected
May 6, 2019
UK cyber-security firm Sophos reported detecting a spike in ransomware attacks at the end of last week from a new strain named MegaCortex. Sophos said the ransomware appears to have been designed to target large enterprise networks as part of carefully planned targeted intrusions –in a tactic that is known as “big-game hunting.” The modus operandi is ...
- Compromised Office 365 Accounts Used to Send 1.5 Million Email Threats in March
May 6, 2019
Microsoft Office 365 remains an attractive target for cybercriminals as it continues to be used by businesses worldwide. In a new report from Barracuda Networks, the company revealed that more than 1.5 million malicious and spam emails were sent from thousands of compromised Office 365 accounts of their customers in March 2019 alone. The increase in the ...
- Mysterious hacker has been selling Windows 0-days to APT groups for three years
May 1, 2019
For the past three years, a mysterious hacker has been selling Windows zero-days to at least three cyber-espionage groups, as well as cyber-crime gangs, researchers from Kaspersky Lab have told ZDNet. The hacker’s activity reinforces recent assessments that some government-backed cyber-espionage groups –also known as APTs (advanced persistent threats)– will regularly buy zero-day exploits from third-party entities, ...
- Dell laptops and computers vulnerable to remote hijacks
May 1, 2019
A vulnerability in the Dell SupportAssist utility exposes Dell laptops and personal computers to a remote attack that can allow hackers to execute code with admin privileges on devices using an older version of this tool and take over users’ systems. Dell has released a patch for this security flaw on April 23; however, many users are likely ...
- APT trends report Q1 2019
April 30, 2019
For just under two years, the Global Research and Analysis Team (GReAT) at Kaspersky Lab has been publishing quarterly summaries of advanced persistent threat (APT) activity. The summaries are based on our threat intelligence research and provide a representative snapshot of what we have published and discussed in greater detail in our private APT reports. ...