YD/T 3953-2021 in English

This standard establishes a technical architecture including five modules: monitoring and filtering, policy generation, user tag analysis, bad message review, and bad message disposal for the converged communication instant messaging service in the IMS network environment. The system achieves a bad message identification accuracy rate of ≥95% through multi-dimensional feature analysis, and the disposal response time is controlled at the millisecond level.

Financial fraud message interception case

After implementing this standard, a certain operator identified the fingerprint of fraud messages with the characteristics of "high-yield financial management" through the feature matching engine, and combined with user social circle analysis to discover abnormal propagation links, blocked the fraud chain involving 50,000 users within 72 hours, and recovered economic losses of more than 10 million yuan.

Main improvements of this standard compared with the 2018 version:
1. Added a fingerprint generation algorithm for rich media messages
2. Introduced a dynamic credit scoring mechanism based on user behavior
3. Established a blacklist sharing mechanism across operators

1. Deployment phase: It is recommended to deploy monitoring probes in the IMS core network bypass first
2. Policy tuning: Dynamically adjust the feature matching threshold according to the actual business volume (the recommended initial value M=0.9, N=0.6)
3. Operation and maintenance: Establish a 7×24-hour hot message tracking mechanism