Name
Title | ||
|---|---|---|
Improved Long Short-Term Memory based anomaly detection with concept drift adaptive method for supporting IoT services |
Abstract | ||
|---|---|---|
The rapid rising of artificial intelligence (AI) and Internet of Things technologies leads to the accumulation of abundant communication data without being processed on time, which causes potential threat for smart city. How to effectively leverage these data for anomaly detection has become an increasingly popular research field as it is a fundamental aspect of cyber security for smart city services. The existing methods often focus on either static data for anomaly detection or streaming data without considering the influence of poor detection accuracy caused by concept drift phenomenon. In this paper, we concentrate on the anomaly detection problem of smart city services and distinguish different anomalies of communication in an effective way, which is aimed to protect data privacy of users. Then we propose an innovative concept drift adaptive method to improve the accuracy of anomaly detection, which fully considers time influence to change the sample distribution along timeline. Furthermore, we present an AI based Improved Long Short-Term Memory (I-LSTM) neural network that adds time factor and employs a novel smooth activation function, which can enhance the performance of multi-classification for anomaly detection. Finally, our proposed methods are evaluated with a real communication dataset. Extensive experimental results indicate that I-LSTM achieves the highest values on all indicators. This demonstrates the effectiveness of our proposed methods that can offer excellent quality of service for smart city, which is a perfect fusion of artificial intelligence and communication security. |
Year | DOI | Venue |
|---|---|---|
2020 | 10.1016/j.future.2020.05.035 | Future Generation Computer Systems |
Keywords | DocType | Volume |
Smart city,Quality of service,Anomaly detection,Concept drift adaptive method,Long Short-Term Memory | Journal | 112 |
ISSN | Citations | PageRank |
0167-739X | 2 | 0.37 |
References | Authors | |
0 | 5 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Rongbin Xu | 1 | 2 | 0.37 |
| Yongliang Cheng | 2 | 10 | 1.83 |
| Zhiqiang Liu | 3 | 2 | 1.05 |
| Ying Xie | 4 | 47 | 14.48 |
| Yun Yang | 5 | 2103 | 150.49 |