Threats Tagged 'ai security'

Cybercriminals are exploiting the recent Claude Code leak incident by using it as a social engineering tactic to deliver malware through GitHub repositories. The attackers have created trojanized versions of the leaked Claude source code, distributing malicious payloads including Vidar stealer version 18.7 and GhostSocks trojan. The campaign demonstrates rapid opportunistic exploitation of high-profile security incidents, with compromised repositories serving as delivery mechanisms. Organizations are advised to implement Zero Trust architecture to mitigate risks from shadow AI instances and trojanized Claude agents. Multiple GitHub repositories have been identified hosting the malicious code, with command and control infrastructure established across multiple IP addresses and domains.

This article analyzes real-world instances of indirect prompt injection (IDPI) attacks targeting AI agents and large language models integrated into web systems. The researchers identify 22 distinct techniques used by attackers to embed malicious prompts in webpages, including visual concealment, obfuscation, and dynamic execution methods. They categorize attacker intents ranging from low-severity disruptions to critical data destruction attempts. Notable findings include the first observed case of AI-based ad review evasion and attempts at search engine optimization manipulation. The article presents a taxonomy of web-based IDPI attacks and provides insights into attack trends based on telemetry data. The researchers emphasize the need for proactive, web-scale defenses to detect IDPI and distinguish between benign and malicious prompts.

MediumCampaignUnited StatesUnited KingdomGermany+8#prompt engineering#ai security#ai agents