• International journal of advanced smart convergence
• /
• v.12 no.3
• /
• pp.51-60
• /
• 2023

This study focuses on the design of a GPT-based system for relatively rapid technology credit assessment of SMEs. This system addresses the limitations of traditional time-consuming evaluation methods and proposes a GPT-based model to comprehensively evaluate the technological capabilities of SMEs. This model fine-tunes the GPT model to perform fast technical credit assessment on SME-specific text data. Also, It presents a system that automates technical credit evaluation of SMEs using GPT and LLM-based chatbot technology. This system relatively shortens the time required for technology credit evaluation of small and medium-sized enterprises compared to existing methods. This model quickly assesses the reliability of the technology in terms of usability of the base model.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2024.01a
• /
• pp.321-324
• /
• 2024

본 논문에서는 LLM(Large Language Models) 기반의 STT 결합 챗봇 시스템을 제안한다. 제조업 공장에서 안전 교육의 부족과 외국인 근로자의 증가는 안전을 중시하는 작업 환경에서 새로운 도전과제로 부상하고 있다. 이에 본 연구는 언어 모델과 음성 인식(Speech-to-Text, STT) 기술을 활용한 혁신적인 챗봇 시스템을 통해 이러한 문제를 해결하고자 한다. 제안된 시스템은 작업자들이 장비 사용 매뉴얼 및 안전 지침을 쉽게 접근하도록 지원하며, 비상 상황에서 신속하고 정확한 대응을 가능하게 한다. 연구 과정에서 LLM은 작업자의 의도를 파악하고, STT 기술은 음성 명령을 효과적으로 처리한다. 실험 결과, 이 시스템은 작업자의 업무 효율성을 증대시키고 언어 장벽을 해소하는데 효과적임이 확인되었다. 본 연구는 제조업 현장에서 작업자의 안전과 업무 효율성 향상에 기여할 것으로 기대된다.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.34 no.3
• /
• pp.451-461
• /
• 2024

Recently, various AI chatbots based on large language models have been released. Chatbots have the advantage of providing users with quick and easy information through interactive prompts, making them useful in various fields such as question answering, writing, and programming. However, a vulnerability in chatbots called "prompt injection attacks" has been proposed. This attack involves injecting instructions into the chatbot to violate predefined guidelines. Such attacks can be critical as they may lead to the leakage of confidential information within large language models or trigger other malicious activities. However, the vulnerability of Korean prompts has not been adequately validated. Therefore, in this paper, we aim to generate malicious Korean prompts and perform attacks on the popular chatbot to analyze their feasibility. To achieve this, we propose a system that automatically generates malicious Korean prompts by analyzing existing prompt injection attacks. Specifically, we focus on generating malicious prompts that induce harmful expressions from large language models and validate their effectiveness in practice.