Malware Trends 2024: AI Rewrites the Rules of Cyber Security

About the job

Artificial Intelligence (AI) will continue to occupy cyber security in the coming year. The range of threats is vast: AI supports cyber criminals in phishing, social engineering, and the development of malware. The greatest concern is that the technology will drive the automation of attacks. Learn more about the predicted malware trends for 2024 in the blog post.

With the success of Chat-GPT, Artificial Intelligence (AI) has finally arrived in everyday life. Generative language models are also valuable tools for cyber criminals. Large Language Models (LLM) generate convincingly human-like texts – which is helpful, for example, in phishing.

The high error rate that previously made recipients suspicious of phishing emails is likely to decrease. This makes cyber criminals' campaigns more effective on the one hand. On the other hand, the gangs' processes also become more efficient. The previously high time expenditure is reduced, as emails can be formulated more quickly and personalised more easily.

Social engineering with the help of deepfakes

AI also brings new risks at the image level. In recent months, various deepfakes have gone viral, demonstrating the astonishing potential of the technology. Fake videos and photos blur the boundaries between reality and fiction. Audio recordings can also be easily faked by putting false statements into real voices. Especially in the government sector, deepfakes, particularly those in real time, can have devastating consequences.

This expands AI's possibilities in social engineering: the technology opens new ways to persuade employees to disclose confidential information or to execute payments. For example, it is conceivable that a superior calls from holiday to arrange an urgent transfer – but on the phone it is not the superior, but an AI with their voice.

AI tools can be easily exploited

However, cyber criminals will not only use AI as a work aid but also deliberately infiltrate AI tools. With the right voice commands, LLM models like Chat-GPT or Bard can be exploited for their own interests. This method is called prompt injection.

The following experiment shows how real this threat is: researchers from Saarland University and the Helmholtz Centre for Information Security (CISPA) succeeded in exploiting Microsoft’s Bing Chat. The research report shows how the seemingly harmless AI assistant was modified into a malicious bot that searches for and collects personal information.

AI drives the automation of attacks

AI can also be helpful in programming malicious software. The technology is likely to be increasingly used in the development of new malware and ransomware variants. This will further fuel the growth of the ransomware-as-a-service industry.

Undoubtedly, the technology will also support the execution of cyber attacks. With the advancement of AI and machine learning, it will increasingly be possible for machines to take over the tasks of hackers. There are great fears that the automation of attacks will make rapid progress.

The good news: AI also supports defence

While cyber criminals are only just discovering AI for themselves, the technology is already being widely used in cyber security. The 2023 study "Artificial Intelligence in Cyber Security" by Mimecast shows: around half of companies in Switzerland and Germany already use AI-supported security solutions – primarily in email security as well as in the detection and defence against threats.

Which other trends will shape cyber security in the coming year? Read our knowledge update "Outlook 2024: The Five Most Important Malware Trends".