From malware creation to network intrusion… AI-based attacks surge over the past year, and zero-day vulnerability discovery models also emerge
);)
The scope of AI misuse is expanding beyond simple hacking attempts to the entire process of an attack. 6.5% of all blocked accounts utilized AI to infiltrate internal networks. The proportion of medium- and high-risk accounts also surged from 33% in the first half of the previous year to 56%, demonstrating that the threat level of AI-based attacks is rapidly increasing.
Actual cases of damage have also occurred. Google announced last month that it had confirmed the first instance of a zero-day vulnerability (undisclosed security flaw) being developed using AI. According to an analysis by Google Threat Intelligence Group (GTIG), attackers implemented a zero-day vulnerability using a Python script that enables the bypassing of two-factor authentication in open-source web-based system management tools. Zero-day vulnerabilities have no defense mechanisms, resulting in a high success rate for attacks, and pose a critical threat to critical infrastructure such as financial systems and virtual asset exchanges.
Antropic is set to soon launch an AI model called 'Mythos.' This model has already discovered over 10,000 critical vulnerabilities in commonly used software. Antropic stated that Mythos independently identified thousands of high-risk vulnerabilities across major operating systems and browsers. While this demonstrates the potential for AI to serve as a security enhancement tool, it also raises growing concerns that it could enable more sophisticated attacks if it falls into the hands of malicious users. Now is the time for virtual asset investors to pay closer attention to strengthening the security of their exchanges and wallets.
Choi Joo-hoon joohoon@blockstreet.co.kr
