Skip to main content

Table 3 Clustering results of various methods

From: Subspace clustering via graph auto-encoder network for unknown encrypted traffic recognition

Dataset

Methods

Input

ACC

NMI

ARI

F1

Ours

K-means

Flow features

0.5122

0.3056

0.3017

0.2898

BIRCH

Flow features

0.4078

0.2678

0.2189

0.2267

GMM

Flow features

0.5478

0.5757

0.3417

0.3516

AE

Flow features

0.5167

0.4165

0.4538

0.3659

Spectral

Flow graph

0.6167

0.5261

0.4770

0.3785

DeepWalk

Flow graph

0.5147

0.3427

0.3618

0.3478

DNGR

Flow graph

0.6787

0.5876

0.5879

0.3897

VGAE

Flow features and Flow graph

0.6784

0.6870

0.6179

0.4157

DAEGC

Flow features and Flow graph

0.7585

0.7658

0.7868

0.5765

SDCN

Flow features and Flow graph

0.8197

0.7998

0.8178

0.6679

SCGAE

Flow features and Flow graph

0.9067

0.8679

0.9235

0.6922

ISCXVPN2016

K-means

Flow features

0.4972

0.3356

0.3416

0.3678

BIRCH

Flow features

0.4256

0.2346

0.3246

0.2290

GMM

Flow features

0.5367

0.5462

0.4471

0.3078

AE

Flow features

0.5435

0.5454

0.4541

0.3512

Spectral

Flow graph

0.5212

0.5263

0.3217

0.3465

DeepWalk

Flow graph

0.5812

0.5168

0.4311

0.3766

DNGR

Flow graph

0.5926

0.5442

0.4443

0.3789

VGAE

Flow features and Flow graph

0.6279

0.6543

0.5214

0.4312

DAEGC

Flow features and Flow graph

0.6927

0.5234

0.5482

0.5891

SDCN

Flow features and Flow graph

0.7403

0.6829

0.6173

0.6612

SCGAE

Flow features and Flow graph

0.8122

0.8124

0.7761

0.7042

  1. Bold text indicates the best experimental results