Social account protection system based on federated spatio-temporal neural network

By utilizing a federated spatiotemporal neural network protection system, which extracts branches in parallel architecture based on spatial and temporal features and combines them with an encryption processing module, the system solves the problems of misjudgment and data leakage in existing protection methods, and achieves efficient social media account protection and privacy protection.

CN122372270APending Publication Date: 2026-07-10CHONGQING COLLEGE OF ELECTRONICS ENG
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHONGQING COLLEGE OF ELECTRONICS ENG
Filing Date
2026-04-15
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing social media account protection methods are insufficient to cope with complex and ever-changing user behavior patterns, and are prone to misjudgment or omissions. Furthermore, centralized artificial intelligence models pose risks of data leakage and insufficient privacy protection.

Method used

A protection system based on federated spatiotemporal neural networks is adopted. Through encrypted communication between cloud server clusters and distributed client devices, a parallel architecture of spatial feature extraction branches and temporal feature extraction branches is used. Combined with a feature fusion layer, a fully connected classification layer and an encryption processing module, the system can achieve spatiotemporal feature analysis of user behavior and privacy protection.

Benefits of technology

It improves the ability to capture features in complex attack scenarios, enhances the depth of identification of covert social media account theft, ensures data privacy and security, and avoids the risk of centralized data aggregation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122372270A_ABST
    Figure CN122372270A_ABST
Patent Text Reader

Abstract

The application discloses a social account protection system based on a federal space-time neural network. A cloud server cluster establishes a bidirectional data connection with multiple client devices through an encrypted communication link. A feature extraction network module comprises a space feature extraction branch and a time feature extraction branch connected in parallel. The output ends of the space feature extraction branch and the time feature extraction branch are connected to a feature fusion layer. The space feature extraction branch and the time feature extraction branch in the feature extraction network module adopt a parallel architecture to independently process the space topology information of geographical coordinates and the time evolution law of operation sequences. The space feature extraction branch extracts local neighborhood features on a two-dimensional grid matrix through convolution operation to identify abnormal jumps of login locations. The time feature extraction branch captures long-term and short-term dependence relationships in a time sequence through a recurrent neural network unit to analyze periodic patterns of user behaviors.
Need to check novelty before this filing date? Find Prior Art