[AI Revolution](44) AI Security Threats Blocked by AI... SK Shieldus

Introduction of AI Technology in Security Monitoring Services
Active Use of AI in Detection Areas
Response to Generative AI Security Threats

"Artificial intelligence (AI) computers like ChatGPT can be exploited for cyberattacks." This is what Sam Altman, CEO of OpenAI, who sparked the generative AI boom with ChatGPT, said in an interview with the US ABC network last March. He added that generative AI is an excellent technology but at the same time a dangerous and frightening entity. Specifically, it can create large-scale false information and write program code that can be used for hacking. If maliciously intended, it can become a deadly weapon above all else. The security specialist company SK Shielders has a 'shield' to block such potential threats. Ironically, the technology that prevents security threats arising from AI is AI technology itself.

SK Shielders introduced AI technology into its 'security monitoring service,' which remotely monitors security systems, analyzes abnormal signs, and determines whether a cyberattack is underway. Security monitoring consists of log collection, detection, analysis, response, and reporting. It operates 24/7, collecting and detecting potential threats, analyzing whether an attack is occurring, and responding by isolating threats through security systems such as firewalls. The 'Secudium' security monitoring platform, developed in 2016, processes approximately 800 million cases annually, 7.9 billion cases daily, and 250,000 cases per second of threat data through big data processing technology.

SK Shieldus Cybersecurity Monitoring Center 'Secudium Center'

For efficient hacking threat assessment, it is explained that AI technology conducts more than 50,000 threat assessments daily. A SK Shielders official said, "If a security monitor assesses security threats directly, they can only handle about 60 cases per day. The logs to be analyzed increase by 15-20% annually, and there is a limitation in that the judgment of whether it is hacking or normal communication varies depending on the individual capabilities of the security monitor. Therefore, AI technology was introduced to improve accuracy and efficiency." AI is actively used especially in the detection area that effectively monitors abnormal traffic. It shows 99.95% accuracy in filtering out 'false positives,' which are cases judged as problematic despite being normal.

SK Shielders is also actively responding to security threats that may arise from generative AI. SK Shielders' white-hat hacker expert group, 'EQST (Experts, Qualified Security Team),' analyzed generative AI attack scenarios mainly into two types: attacks targeting AI models and training data, and threats arising from abusing AI-utilized services. Threats targeting AI models include manipulating input data or adding malicious data to manipulate the model. 'Prompt injection' attacks, which bypass guidelines or policies applied within AI services through malicious queries to extract responses, are a way of abusing AI-utilized services. In fact, prompt injection occurred in multiple plugins introduced by OpenAI in March this year. Lee Jae-woo, head of SK Shielders EQST Business Group, said, "As generative AI advances, automation of attacks and leveling up of attack sophistication will occur. Generative AI will be essential on the defense side to respond to attacks."

SK Shielders plans to continue sharing safe AI usage methods by presenting AI checklists from the perspectives of users, developers, and companies. The group leader said, "EQST will research the application of generative AI on the defense side and proactively propose countermeasures against increasingly sophisticated attacks."

This content was produced with the assistance of AI translation services.