A Supervised Learning Approach: Detection of Cyber Attacks

Abstract

Modern technology has brought the Internet of things (IoT) which is a blessing for the nearly 8 billion people in the world. Using the advancement of IoT, we need not stay at our home all the time to use our appliances because IoT gives us a choice to use them from anywhere. Transferring and receiving data from one device to another becomes too easy with the remote monitoring process as IoT connects all the devices to the internet. But IoT infrastructure can be affected by a couple of different attacks and anomalies as it uses IoT sensors and wireless devices. As the public shares lots of their confidential and private data, it is necessary to establish user security and privacy by detecting intrusions and malware in this infrastructure. In this paper, 5 different supervised machine learning algorithms K Nearest Neighbors (KNN), Naive Bayes, Support Vector Machine (SVM), Random Forest, and Decision Tree Classifier have been used to detect attacks in different computer networks which are listed in CICIDS 2017 dataset. The paper shows the novel approach of detecting new attacks by extracting the highest weight scored 25 features using Random Forest Regressor and Extra Tree Classifier to analyze different cyberattacks by implementing different supervised learning models. After performing a comparison analysis between the 5 algorithms the paper finds that the KNN model performs better than others by giving the highest F1 score and accuracy.