A Method for Identifying Tor Users Visiting Websites Based on Frequency Domain Fingerprinting of Network Traffic

Table 1

Well-performing website fingerprinting methods.

Method	Accuracy (%)	Advantage	Disadvantage
K-NN [7]	95.0	Multiple features are used.	The noise is not considered.
CUMUL [8]	97.3	The noise is considered.	Overfitting.
K-FP [9]	95.5	Sorting time is shortened, and overfitting is reduced.	Classification effect is easily affected by the noise.
DF [15]	98.3	Complex CNN model is proposed.	Large datasets are needed.
TF [16]	95.0	Small datasets are needed.	The accuracy needs to be improved.
Tik-Tok [18]	98.4	Multiple time features are considered.	Classification effect is easily affected by circuit congestion.