Undergraduate students from the Department of AI and Big Data Convergence Management at Kookmin University presented a research paper at ICPR 2026 (International Conference on Pattern Recognition), a prestigious international academic conference in the field of artificial intelligence, held in Lyon, France, from August 17 to 22.

The paper presented at this conference was titled “Efficient Korean Voice-Phishing Detection using QLoRA-tuned Small Language Models,” authored by students Koo Jun-hoe and Park Eun-woo from the Department of AI and Big Data Convergence Management, under the guidance of Professor Lee Jae Hyuk. The study proposed a method for utilizing language models to effectively detect voice phishing, a crime that has recently caused a sharp increase in social harm.

Instead of using large language models (LLMs), which require high computational costs and long inference times, the research team built a voice phishing detection model by fine-tuning publicly available small language models (SLMs). In particular, they applied the QLoRA (Quantized Low-Rank Adaptation) technique to enable efficient model training even with limited computing resources, and utilized a back-translation-based data augmentation technique to ensure the diversity of the training data. Through this approach, they confirmed that it is possible to improve voice phishing detection performance and reduce inference time while using relatively small-scale language models. This research is particularly significant because it demonstrates the potential for application in real-time voice phishing detection systems by considering not only detection accuracy but also computational efficiency.

Professor Lee Jae Hyuk stated, “This research is particularly significant because it was achieved through the proactive participation of undergraduate students, from defining the research topic to developing the model and conducting experiments.” He added, “It is also significant in that it presents a practical solution using AI technology to the problem of voice phishing, which is causing serious harm in society.” He continued, “Going forward, the Department of AI and Big Data Convergence Management will continue to foster a culture where students can take the lead in research and will actively support them in developing AI technologies that can be directly applied to various real-world problems.”