An AI-based computer abnormal traffic detection system
By combining hypersphere feature space partitioning and sparsity modulation mechanism with prior adjustment factor and composite chaotic mapping strategy, the problem of difficult identification of sparse attack features in the existing technology is solved, realizing high-precision detection of covert attack traffic and reducing false alarm rate and false negative rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANTONG INST OF TECH
- Filing Date
- 2026-06-15
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies lack mechanisms for spatiotemporal joint extraction of statistical and load features and feature modulation through spatial sparsity. This makes it difficult to effectively strip away disguises when facing covert attack traffic. Furthermore, the feature filtering method from a global perspective allows the distribution pattern of normal traffic to mask the weak features of sparse attacks. The lack of a targeted dimensionality reduction intervention mechanism to amplify the features of sparse attacks results in extremely low recognition.
The feature extraction module obtains spatiotemporal joint feature vectors through hypersphere feature space partitioning and spatial sparsity modulation mechanism. Combined with the feature dimensionality reduction module, a core feature index table is constructed using prior adjustment factors and global information gain rate. Support vector machine is used for traffic anomaly detection. The parameters are optimized using composite chaotic mapping and natural nearest neighbor density evaluation strategy to achieve high-precision classification of sparse attacks.
It effectively strips away high-risk traffic disguised as normal business traffic, amplifies the subtle characteristics of sparse attacks, reduces the false alarm rate and false negative rate of security devices in complex network environments, and achieves high-precision decision boundary fitting for sparse attack traffic clusters.
Smart Images
Figure CN122419984A_ABST