Comparing and Analyzing Applications of Intelligent Techniques in Cyberattack Detection

Table 4

Performance comparison of different machine learning techniques.

Parameters	ANN	NB	SVM	KNN	DT	RF	DL
Accuracy	High	High	High	Low	Low	High	High
Training time	High training time due to complexity	Low training time due to simplicity	High training	High training time	High training time due to complex structuring	High training time	High training time complex structuring
Execution time	Average	High	High	Low	Low	Low	High
Large attributes	Dealing well with large attributes	Dealing well with large attributes	Dealing good with large attributes but speed will be very slow	Dealing well with large attributes	Average dealing with larger attributes	Dealing well with large attributes	Dealing good with large attributes but speed will be very slow
Lots of missing attributes	Contradictory	Good performed	Good performed	Low performed	Good performed	Good performed	Good performed
Lots of noisy data	Contradictory	Better dealing with noise	Better dealing with noise	Low capability dealing with noisy data	Average dealing with noise	Average dealing with noise	Better dealing with noise but the overall process is time-consuming
Large datasets	Cannot handle large dataset and speed of processing will be very slow	Better while handling large dataset	Average performed while handling large dataset but processing speed will be very slow	Better while handling large dataset	Average performed	Average performed	Better while handling large dataset but processing speed will be very slow due to complex structure
Detection rate	High	High	High	Low	High	Low	High
Datasets suitable	KDD99 and NSL-KDD	KDD99 and NSL-KDD	NSL-KDD, KDDCUP99, and DARPA	NSL-KDD, KDDCUP99, and DARPA	KDDCUP99	KDDCUP99	KDD99 and NSL-KDD