TY - GEN
T1 - Anomaly Detection in Software Defined Networks Using Ensemble Learning
AU - Krzemień, W.
AU - Jędrasiak, K.
AU - Nawrat, A.
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - The goal of the article was the detection of anomalies and attacks in Software Defined Networks (SDN) with the appropriate selection and use of artificial intelligence algorithms and their parameters. The main challenge of the work was to determine which ensemble learning algorithm and with what parameters will allow the best detection of attacks and to compare these results with the detection using neural networks. The research was carried out on six sets using the methods of ensemble learning, i.e. RandomForestClassifier, ExtraTreeClassifier, AdaBoostClassifier, GradientBoostingClassifier and XgBoost. The RandomForest and xgboost algorithms turned out to be the most effective in detection, while the largest compromise between the detection speed and the effectiveness of RandomForest. The percentage of detected anomalies on the tested data oscillated at the level of 99–100%, which is a satisfactory result. The innovation in the approach to the problem is that the algorithm works efficiently independently of the input dataset, as opposed to the current solutions, where the algorithm is adapted to a dedicated input set.
AB - The goal of the article was the detection of anomalies and attacks in Software Defined Networks (SDN) with the appropriate selection and use of artificial intelligence algorithms and their parameters. The main challenge of the work was to determine which ensemble learning algorithm and with what parameters will allow the best detection of attacks and to compare these results with the detection using neural networks. The research was carried out on six sets using the methods of ensemble learning, i.e. RandomForestClassifier, ExtraTreeClassifier, AdaBoostClassifier, GradientBoostingClassifier and XgBoost. The RandomForest and xgboost algorithms turned out to be the most effective in detection, while the largest compromise between the detection speed and the effectiveness of RandomForest. The percentage of detected anomalies on the tested data oscillated at the level of 99–100%, which is a satisfactory result. The innovation in the approach to the problem is that the algorithm works efficiently independently of the input dataset, as opposed to the current solutions, where the algorithm is adapted to a dedicated input set.
KW - Anomaly detection
KW - Artificial intelligence
KW - Ensemble learning
KW - SDN
KW - Software-defined networks
UR - https://www.scopus.com/pages/publications/85126988839
U2 - 10.1007/978-3-030-98015-3_44
DO - 10.1007/978-3-030-98015-3_44
M3 - Conference contribution
AN - SCOPUS:85126988839
SN - 9783030980146
T3 - Lecture Notes in Networks and Systems
SP - 629
EP - 643
BT - Advances in Information and Communication - Proceedings of the 2022 Future of Information and Communication Conference, FICC
A2 - Arai, Kohei
PB - Springer Science and Business Media Deutschland GmbH
T2 - Future of Information and Communication Conference, FICC 2022
Y2 - 3 March 2022 through 4 March 2022
ER -