Accurate Prediction and Detection of Adversarial Attack Through CatBoost Library

Aim: The fundamental objective of the research is to use the CatBoost library to increase the accuracy with which adversarial assaults are detected in wireless networks.. In an effort to get higher accuracy, the performance of CatBoost is tested against well-known algorithms like Gradient Boosting, K-nearest neighbor, Random Forest, and Support Vector Machine (SVM). Materials and Methods: There are two groups in the study. Group 1 concentrates on applying classical machine learning algorithms, which have shown comparatively lower accuracy rates, such as SVM, Random Forest, K-nearest neighbour, and Gradient Boosting. Group 2 focuses on accurately identifying adversarial attacks using CatBoost. The technique makes use of a preprocessedDDos SDN dataset to promote efficient intervention and further this field's investigation. Results: With respect to accurate adversarial attack analysis, the suggested method performs exceptionally well, attaining excellent performance metrics including F1 score (0.98), recall (0.96), precision (0.97), and accuracy (0.98). These metrics highlight how useful the suggested detection analysis method is. Conclusion: The efficiency of CatBoost in identifying adversarial attacks is demonstrated in the study's conclusion, underscoring its superiority over conventional techniques like Extreme Gradient Boosting.