|
|
Bimonthly Since 1986 |
ISSN 1004-9037
|
|
|
|
|
Publication Details |
Edited by: Editorial Board of Journal of Data Acquisition and Processing
P.O. Box 2704, Beijing 100190, P.R. China
Sponsored by: Institute of Computing Technology, CAS & China Computer Federation
Undertaken by: Institute of Computing Technology, CAS
Published by: SCIENCE PRESS, BEIJING, CHINA
Distributed by:
China: All Local Post Offices
|
|
|
|
|
|
|
|
|
|
Abstract
A machine-learning-based IDS is crucial for protecting our economy and military in light of the increasing global interconnection of internet infrastructures and the enormous amounts of data being produced on a daily basis. For intrusion detection, researchers employ a unified learning model that draws from both classical and modern deep learning techniques. The one-learning-model approach might fail if data distributions and incursion patterns become too complicated. It's possible that a single deep learning model won't be sufficient to properly capture distinguishing patterns for intrusive attacks with little data. With the goal of further enhancing the performance of machine learning-based IDS, we present the Big-Data-based Hierarchical Deep-Learning System (BDHDLS). BDHDLS employs behavioural features and content features to comprehend the intricacies of network traffic and the information contained in the payload. Each deep learning model in BDHDLS is trained using data from a single cluster in mind. As compared to approaches that used a single learning model, this one has the potential to increase the frequency with which hostile incursions are discovered. Because of its parallel training approach and massive data capabilities, BDHDLS significantly reduces the time required to create a model when several machines are employed.
Keyword
#
PDF Download (click here)
|
|
|
|
|