This article discusses the growing threat of AI-powered cyberattacks by presenting five examples:
-
AI-Powered Fraud (Deepfakes): Generative AI can create realistic audio and video deepfakes by mimicking a person's voice or appearance. These can be used to impersonate individuals and trick employees into transferring large sums of money, as seen in real-world examples. Even short audio clips can be enough to create a convincing deepfake.
-
AI-Powered Phishing: AI significantly enhances phishing attacks by generating emails with perfect grammar and personalized content, making them harder to identify as fake. LLMs can create convincing phishing messages quickly and effectively, even for attackers who don't master the language. This bypasses traditional clues like poor spelling and grammar.
-
AI-Powered Login Attacks: AI agents, like "brute force AI," use Large Language Models (LLMs) to identify login pages and then attempt to gain access through brute-force or password spraying techniques. The AI automates the process, making it easier for attackers to find vulnerabilities.
-
AI-Based Ransomware: Projects like "prompt lock" demonstrate how AI agents, powered by LLMs, can plan and execute ransomware attacks. They analyze systems, identify sensitive data, determine ransom amounts, generate attack code, and even write personalized ransom notes. This can lead to polymorphic attacks that are difficult to detect and ransomware-as-a-service offerings.
-
AI-Powered Exploits: AI can automate the process of finding and exploiting vulnerabilities. Tools like "CVE genie" can take publicly available vulnerability information (CVEs) and use LLMs to understand the vulnerability and generate the actual exploit code. This lowers the technical bar for attackers, enabling them to launch sophisticated attacks with minimal coding knowledge and at a very low cost. This also extends to generating polymorphic malware.
Defenders will need to leverage AI for prevention, detection, and response to counter these evolving threats.