Generative artificial intelligence (GenAI) services are already being leveraged by cybercriminals to create convincing phishing emails, and it appears that these tools are being used for the creation of malware. GenAI services are capable of writing code; however, guardrails have been implemented to prevent malicious uses of these tools, such as the creation of malware. If those guardrails can be circumvented, the creation of malware would no longer be limited to skilled malware developers. Lower-skilled cybercriminals could develop their own malware using GenAI services, and there is growing evidence they are doing just that.
Over the summer, HP security researchers identified an email campaign targeting French users. The phishing email used HTML smuggling (encrypted HTML) to evade detection, and on analysis, the campaign delivered malicious VBScript and JavaScript code that appeared to have been created using GenAI tools. The entire malicious code included comments about what each function does, which is rare in malware development as the exact workings of the code tend not to be described. The comments, along with the use of native language function names and variables all suggest that GenAI was used to create the malware.
The code was used to deliver AsyncRAT malware, a widely available, open source malware that is an information stealer capable of recording the victim’s screen and logging keystrokes. The malware also acts as a malware downloader that can deliver other malware payloads, including ransomware. In this campaign, little technical skill was required as HTML smuggling does not require any programming, the malware being delivered is widely available, and the fact that the comments had not been removed and there was no obfuscation, points to the development of malware by an inexperienced cybercriminal.
There have been other examples of apparent malicious code creation using GenAI, such as a malicious PowerShell script identified earlier this year that was also used to deploy infostealer malware. That campaign targeted users in Germany and impersonated Metro cash-and-carry and was also delivered via email. Just as GenAI tools are helping writers rapidly create written content, GenAI tools can be used to rapidly develop malicious code. ChatGPT and Gemini have guardrails in place that it may be possible to circumvent, but there are many dark LLMs that lack those controls such as WormGPT and FraudGPT. If these tools are leveraged, relatively low-skilled cybercriminals can develop their own malware variants.
Traditional antivirus solutions use signature-based detection. When malware is identified, a signature is added to the antivirus solution for that specific malware variant that allows it to be detected in the future. There is a delay between the creation of malware and the addition of malware signatures to the definition lists of antivirus solutions, during which time malware can easily be smuggled onto devices undetected. If the creation of malware can be accelerated with GenAI tools, cybercriminals will have the upper hand.
The solution for businesses is to deploy security solutions capable of detecting novel malware variants by their behavior rather than a signature. Since malware is commonly delivered via email, having a cloud-based email security solution that incorporates behavioral analysis of attachments will help identify and neutralize these malware variants before they can be installed.
SpamTitan from TitanHQ is a cloud-based antispam software that incorporates email sandboxing. When standard antivirus checks are passed, suspicious emails and attachments are sent to a next-generation email sandbox for deep inspection, where the behavior of the attachments is assessed in an isolated sandbox environment. If malicious actions are detected, the threat is neutralized. SpamTitan also incorporates AI-based and machine-learning detection mechanisms to assist with malicious email detection, and along with a host of other checks ensure malicious emails are detected and blocked. In recent independent tests, SpamTitan has a 99.99% phishing catch rate and a 100% malware catch rate, with zero false positives.
SpamTitan, like all other TitanHQ cybersecurity solutions, is available on a free trial to allow you to see for yourself the difference it makes. To find out more about protecting your business from increasingly sophisticated threats, give the TitanHQ team a call.