SK shieldus was recognized for research competitiveness in the global AI academic community by leveraging artificial intelligence (AI) security technology that can analyze irregular cyberattack patterns.
SK shieldus said on the 22nd that a research paper by Senior Researcher Lim Jeong-hoon from its in-house cybersecurity AI research organization, "Cybersecurity AI Labs," was accepted to ICML 2026, one of the world's three major AI conferences. ICML is a leading academic event, along with NeurIPS and ICLR, that is counted among the world's three major AI conferences.
Conventional cybersecurity detection technologies generally analyzed data on the assumption that attacks proceed in a consistent sequence. In reality, however, cyberattacks occur at irregular times and intervals, and they appear in various forms, such as concentrating in a short period or unfolding over a long span. As a result, it was difficult to sufficiently reflect changing patterns with existing methods, leading to missed signs of attack or reduced detection accuracy.
In response, SK shieldus proposed an AI technology, "QuITE (Query-based Irregular Time-series Embedding)," that can directly analyze attack sequences that unfold irregularly.
QuITE is an analysis technique that effectively represents data with varying time intervals. It is designed to more naturally reflect real attack flows and can flexibly integrate with existing AI models, giving it scalability for application across diverse security detection systems. In addition, QuITE showed up to a 45.9% performance improvement over conventional time-series analysis methods on global public benchmark datasets.
Senior Researcher Lim Jeong-hoon of SK shieldus said, "In the AI academic community, how to effectively handle imperfect data from real-world environments is being discussed as an important task," and noted, "This study is meaningful in that it enables existing AI models to learn irregular attack patterns with greater precision."
SK shieldus is reviewing applying this research outcome to major security services such as its cybersecurity monitoring center "Secudium" and "MDR." If applied, the company expects to improve precision across the entire process and identify anomalous signs that were difficult to capture with conventional methods. The research will be presented to AI researchers from around the world at ICML 2026, to be held at COEX in Seoul from July 6.
Kim Byung-mu, executive vice president of the Cybersecurity Division at SK shieldus, said, "With the acceptance of this paper at ICML, SK shieldus' AI research capabilities have been recognized as competitive in the global academic community," and added, "We will link the research outcomes to our services to elevate detection and analysis capabilities to the next level, and we will continue to expand investment and research in AI technologies specialized for cybersecurity."