PROJECT & FUNDING SCHEMES

According to data from HKCERT, phishing cases in Hong Kong surge 108% from 3,752 in 2023 to 7,811 in 2024, with smishing (SMS phishing) becoming especially common due to high mobile and e-commerce use. Existing protection measures depend on users installing message-filtering apps after delivery, though many refrain from doing so due to privacy concerns. On the other hand, there lacks detection models tailored to Hong Kongstyle SMS due to the scarcity of Chinese SMS training data.  To address these challenges and strengthen Hong Kong's overall cyber resilience, a proactive phishing detection system that identifies and mitigates SMS-based fraud in real time is proposed. Rather than depending primarily on end-user awareness or after-thefact filtering, the proposed solution enables early interception and detection of malicious messages at the point of delivery. By integrating advanced neural network models with text feature analysis leveraging large language model’s ability, the system aims to deliver more accurate, efficient, and contextually relevant protection against rapidly evolving smishing threats, tailored to the specific patterns and linguistic characteristics prevalent in Hong Kong.
With the proposed system, users can benefit from real-time, locally tailored smishing detection, thereby strengthening cyber resilience across the board and minimizing active end-user involvement.