ld258166011/VAEN

Traffic Identification of IMTD17 Dataset Based on Variational Autoencoder Network.

Variational Autoencoder Network

Variational Autoencoder Network (VAEN) implementation on IMTD17 datasets to realize traffic identification.

Introduction

3 files are available in this repository:

File Name	Description
model.py	Variational Autoencoder Network
figure.py	Figure drawing helper functions
main.py	Main process in TensorFlow

Note: .py files are written for Python 3.x.

Application

Sample Reconstruction

Sample reconstruction of (a)Alipay, (b)Kugou, (c)Weibo traffic image.

Latent Representation

2-D Latent representation distribution.

Note: Labels are corresponding to IMTD17 datasets

Sample reconstruction from 2-D latent representation.

Traffic Identification

Classification result under different threshold. Classification Result of App Traffic.

App	Pre	Rec	App	Pre	Rec
Alipay	100	100	QQ	100	100
Baidu	100	100	QQ Mail	100	93.2
Bilibili	100	100	QQ Music	100	100
CNTV	100	100	Taobao	100	100
JD	100	89.1	WeChat	100	86.6
Kugou	100	94.6	Weibo	100	100

Note: Pre for Precision, Rec for Recall