- Abstract
背景:技术背景以及当前需求;
解决方案:基于深度学习对网络流量进行特征提取;
本方案优点:实时优化计算资源、微调检测与分析行为;
实验证明:对于上述优点的佐证。
- 简介(INTRODUCTION AND MOTIVATION)
现状以及当前方案的缺点:数据量、设备数量、网速、低延时等特点导致现有方案失效;
提出方案:
实验验证:
文章结构:
- 相关技术当前在相关领域中的应用
挑战
现有相关方法:
- DEEP LEARNING APPLIED TO THE ANOMALY DETECTION PROBLEM IN 5G NETWORKS
自编码器:
SAE算法:
https://blog.csdn.net/xiatianyunzi/article/details/82456125
https://blog.csdn.net/llh_1178/article/details/80274468
DBN算法:
https://blog.csdn.net/kellyroslyn/article/details/82668733
https://blog.csdn.net/u011501388/article/details/78202093
-
实验结果
-
结论和展望
-
专有名词:
ASD:Anomaly Symptom Detection
NAD:Network Anomaly Detection
VI:Virtualized Infrastructure
VNF:Virtualized Network Functions
DPI:Deep Packet Inspection
EPC: Evolved Packet Core (EPC)
IDS:Intrusion Detection Systems
VNO:Virtual Network Operators
Key Performance Indicators (KPI)
TP:True Positive
TN:True Negative
FP:False Positive (FP).
FN:False Negative (FN).
设备:
EPC:Evolved Packet Core
UE:User Equipments (UE)
RAN:Radio Access Network
算法:
BBNN: Block-Based Neural Network (BBNN)
DBN:Deep Belief Networks
SAE:Stacked AutoEncoders
LSTM:Long Short-Term Memory Recurrent Networks
RBM:Restricted Boltzmann Machine (RBM)
SVM:Support Vector Machine