Analysts Cyberk Labs described How to create polymorphic malicious programs using ChatGPT. Researchers plan to publish part of their work “for educational purposes”.
Experts went around ChatGPT content filters and showed how “with minimal efforts, you can continuously request a unique, functional and proven fragment of code in ChatGPT. The attacker can create polymorphic malicious software, which is not detected by means of protection and leaves no traces of memory.
To bypass the content filter, analysts have “persuaded” ChatGPT write the program code by describing their request in detail. The chatbot created part of the DLL (DLL Inject) code into the “Explorer.exe” in Windows and encode it in Base64. Then the researchers using ChatGPT created the shell-code injector.
Part of the generated code
The use of ChatGPT in creating malicious software gives an attacker with minimal efforts to create a reliable exploit for 1 day, which will cause alarm from suppliers of security solutions, because at the moment the development of protective software is not concentrated on the malicious ones created by Chatgpt.