Using clustering to improve the KNN-based classifiers for online anomaly network traffic identification

A pre-assessment of past research on the topic of environmental-friendly electronics

“…For any two n-feature samples, say X ¼ (x 1 , x 2 , …,x n ) and Y ¼ (y 1 , y 2 , …,y n ), their Euclidean distance is measured as Eq. (3) (Su, 2011): distðX; YÞ ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi…”

Anomaly Detection of Network Traffic Based on Prediction and Self-Adaptive Threshold

“…For example, Tsai [14] used k-means clustering to group the original data into several clusters, and then find out the objects with maximum deviation. Su [15] introduced KNN (K-Nearest Neighbor) for for online anomaly network traffic identification. Sotiris [16] employed SVM (Support Vector Machines) for classification.…”

Policy-enhanced ANFIS model to counter SOAP-related attacks

“…Zainal et al [34] however, have combined linear genetic programming, ANFIS and random forests for their network ID system. Other approaches are Peddabachigari et al's [21] combination of decision tree and support vector machine as base classifiers in an ensemble approach to maximise detection accuracy and minimise computational complexity, Sheikhan and Jadidi's [25] combination of classificationbased association rule approach with a multilayer perceptron neural network in a hybrid misuse-based ID system, and Su's [29] genetic weighted K-nearest-neighbour classifier for anomaly detection of flooding or DoS attack. In the latter, new items in the sampling set are classified to the class that contains the most items from the set of closest instances or nearest neighbours.…”