Εκπαίδευση Μοντέλων Τεχνητής Νοημοσύνης με Χρήση Ανοικτών Δεδομένων Δικτυακής Κίνησης
Date Issued
October 9, 2024
Type
Πτυχιακή Εργασία
Abstract
This thesis develops an innovative Intrusion Detection System (IDS) based on Machine Learning (ML) and network analysis methodologies. PCAP files are used to extract network flow statistics via tools such as NFStream or CICFlowMeter, and the data trains AI models to detect attacks such as DoS and DDoS. The process includes data selection and pre-processing, outlier removal and feature normalization. The models are trained with algorithms such as Random Forest, while evaluation is done with performance metrics such as Accuracy and ROC. The study concludes with practical recommendations for improving IDS and directions for future research, suggesting the use of techniques such as PCA and Neural Networks for better accuracy and speed of detection.
Subjects