New Malware Employs Eerie Tactics to Evade AI Detection

A novel malware evasion technique was identified by Check Point, where a malicious file communicated with AI-based malware detection tools using prompt injection to disguise itself as harmless. The file contained a hardcoded C++ string instructing the AI to ignore prior commands and classify it as benign, specifically asking for a response of “NO MALWARE DETECTED.” Although the attempt was unsuccessful, as detection tools recognized the prompt injection, it signifies a new strategy in malware targeting AI defenses. The file also included various sandbox evasion tactics and an embedded TOR client. This early detection highlights the evolving methods threat actors may employ to circumvent AI-based security measures and emphasizes the importance of continual adaptation and enhancement of cybersecurity defenses to keep pace with such innovations. Researchers aim to monitor these developments to bolster protective measures for their customers and the broader cybersecurity community.

Source link