Machine Learning-based Detection of Web Attacks Using Logistic Model Trees

Cyberattacks pose a significant threat to internet security and may cause great damage to businesses. Machine learning methodologies have been widely used to detect cyber intrusions. This study evaluated the effectiveness of two machine learning techniques, smooth private forest and logistic model trees, in identifying cyber breaches using the CICIDS 2017 dataset. The findings indicated that both models, smooth private forest and logistic model trees, attained high accuracy (99.54%), positive accuracy (99.50%), retrieval (99.50%), and F1 score (99.40%). The SPF paradigm has robust privacy safeguards, making it appropriate for applications where data confidentiality is paramount. The examination of the importance of features in the study indicates that Fwd Packet Length Mean, Fwd IAT Min, and Init_Win_bytes_backward are critical for detecting web intrusions. These discoveries may facilitate the development of more efficient intrusion detection systems for online security. This research illustrates the efficacy of smooth private forest and logistic model tree models in online attack detection and their significance in enhancing web security.