SK Shieldus, a leading security service company in Korea, warned about the increasing cyberattacks using generative artificial intelligence targeting new technology vulnerabilities. At a media seminar, SK Shieldus and its white-hat hackers group, EQST, highlighted the importance of awareness and preventive measures to mitigate damage from these advanced cyber threats. Large Language Models are aiding hackers in sophisticated attacks, with deepfake phishing and virtual asset theft being major concerns globally. In Korea, most cyber infringement cases occurred in the finance industry, followed by IT and communication and manufacturing sectors. Globally, public and administration sectors were most affected by cyberattacks. EQST identified three critical vulnerabilities related to AI fraud and hacking using generative AI and chatbots, including prompt injection, insecure output handling, and sensitive information disclosure. Companies are now customizing small language models to train only necessary information for their work. To prevent such attacks, SK Shieldus recommended prompt security and data filtering solutions.

Source link

Source link: https://m.koreaherald.com/amp/view.php?ud=20240702050602