Network attack behavior detection based on double randomization design of random forest algorithm

Aiming at the problem of the traditional network security defense system in the detection of network attack behavior, this paper proposes an active detection model of network attack behavior based on random forest algorithm. By constructing a dual-layer decision architecture and integrating MapReduce distributed computing framework and dual randomized integrated learning mechanism, network traffic features can be efficiently extracted and dynamic threat perception can be realized. The model adopts multi-modal classifier coordination strategy, combined with data local optimization, heterogeneous decision tree integration and soft voting fusion technology, which breaks through the limitations of traditional detection methods in feature correlation modeling and noise robustness. Experimental results show that compared with Adaboost algorithm, the proposed model has significant advantages in detection rate (DR), Accuracy and Precision, which verifies the stability and generalization ability of random forest algorithm in network attack detection tasks.