Comparing and Analyzing Applications of Intelligent Techniques in Cyberattack Detection

Table 3

Comparative study among different machine learning techniques.

Technique	Principle	Parameters	Advantages	Limitations
k-means	Find out k points called centers that are evaluated as the sum of the distances of all points to their respective cluster centers	Cluster center location	High computation, produce closet clusters	Calculation of K is a very tough task for a fixed number of clusters. The dissimilarity, the initial and final cluster partition
K-nearest neighbor	The input consists of k-closest training of the feature space by using instance-based learning	Class of nearest neighbor	It is easy to implement, less complex	Difficult to deal with arbitrary attributes
Support vector machine	The mapping of input data to the high-dimension space and also dealing with linearly separated data for classification	Features of high dimension	Having high accuracy, flexible and robust in dealing with errors	Takes large time for training, complex to handle learned function (weights)
Hidden Markov model	It is a statistical or sequence-based model that consists set of states, transitions represent the set of possible positions	Pixels in a vision-based input	High-scalable model and easy to understand	Many assumptions about the data. A large number of parameters required to be set. Highly needed training data