Cloud-Enhanced Intrusion Detection : Evaluating Deep Learning and Ensemble Methods Using Hikari 2021 Dataset
Sohail, Razi (2024)
2024
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2024052214414
https://urn.fi/URN:NBN:fi:amk-2024052214414
Tiivistelmä
The thesis provides an in-depth analysis into machine learning and deep learning system methodologies for network invasion discovery using the HIKARI-2021 dataset. This involved the application of models like Random Forest, XG Boost, LSTM, and GRU in identifying and classifying various malicious undertakings transmitted within the network exchange.
Depending on the operation, the models were accessed primarily with reference to their accuracies and their confusion matrix evaluations. The results of this study indicated Random Forest with an accuracy of 93.77%, XG Boost at 93.02%, LSTM at 92.48%, and GRU at 92.50%. Subsequently, this was related to benchmark system models as demonstrated in related literature contrast with random performances of between 98% and 99% accuracy.
This research concludes that, based on form of entry employed, the model is imminently viable because each has its power, weakness, and potential in real-earth usage. Additionally, there is an opportunity for optimizing these system models and designing new features based on the lessons learned from this analysis.
Depending on the operation, the models were accessed primarily with reference to their accuracies and their confusion matrix evaluations. The results of this study indicated Random Forest with an accuracy of 93.77%, XG Boost at 93.02%, LSTM at 92.48%, and GRU at 92.50%. Subsequently, this was related to benchmark system models as demonstrated in related literature contrast with random performances of between 98% and 99% accuracy.
This research concludes that, based on form of entry employed, the model is imminently viable because each has its power, weakness, and potential in real-earth usage. Additionally, there is an opportunity for optimizing these system models and designing new features based on the lessons learned from this analysis.