ElNnGA: An Artificial Intelligence Empowered Scheme for Unknown Attack Recognition and Defense Strategy Recommendation

Jiale Quan, Yu Han, Liumei Zhang, Yuanjiao Hu

Abstract: With the diverse forms of current network attacks, attackers continue to evolve and improve their attack methods, making traditional security countermeasures gradually lag behind. This study is dedicated to solving the problem of identifying unknown attacks and recommending defense strategies, and proposes an artificial intelligence-based scheme. First, the data processing is accomplished by self-partitioning the UNSW-NB15 dataset into test and training sets as well as feature reduction, which ensures the balance of the classification task, followed by the integration of Random Forest and XGBoost classifiers to realize the binary classification of network traffic data. Next, feed-forward neural networks are utilized to identify the nine attack types of malicious data for multi-classification. After identifying the specific attack types, genetic algorithms are used to calculate the data of defense cost, defense efficacy and benefit of various attack types, so as to iterate the optimal defense strategy recommendation. Experimental results show that the classification algorithm and genetic algorithm selected in this paper have good performance and adaptability for unknown attack identification and defense strategy recommendation, compared with the traditional classification model, this study not only bicategorizes unknown attacks, but also identifies the specific attack types, with high classification accuracy, and the design of the fitness selection function effectively improves the convergence of the genetic algorithm, which provides important recommendations for the selection of defense strategies for different attacks. strategy selection, which provides an important recommendation guidance for different attacks.

External IDs:dblp:conf/nana/QuanHZH24