Development of intrusion detection system using residual feedforward neural network algorithm
An intrusion detection system (IDS) is required to protect data from security threats that infiltrate unwanted information via a regular channel, both during storage and transmission. This detection system must differentiate between normal data and abnormal or hacker-generated data. Additionally, th...
Saved in:
Main Authors: | , , , , , |
---|---|
Format: | Conference or Workshop Item |
Language: | English English |
Published: |
IEEE
2022
|
Subjects: | |
Online Access: | http://irep.iium.edu.my/96797/2/2021_ISRITI_Cert_Development_of_Intrusion_Detection_System_using.pdf http://irep.iium.edu.my/96797/13/96797_Development%20of%20intrusion%20detection%20system.pdf http://irep.iium.edu.my/96797/ https://edas.info/web/20214thisriti/index.html https://doi.org/10.1109/ISRITI54043.2021.9702773 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | An intrusion detection system (IDS) is required to protect data from security threats that infiltrate unwanted information via a regular channel, both during storage and transmission. This detection system must differentiate between normal data and abnormal or hacker-generated data. Additionally, the intrusion detection system (IDS) must be precise and quick to analyze real-time traffic data. Despite extensive research, there is still a need to improve detection accuracy and speed due to the tremendous increase in internet traffic volume and variety. This paper introduces a novel, efficient, and accurate approach for real-time intrusion detection and classification based on the Residual Feedforward Neural Network (RFNN) algorithm. The RFNN algorithm is developed to avoid overfitting, improve detection accuracy, and accelerate training and inference. Additionally, the suggested algorithm is highly adaptable and straightforward to accommodate different types of intrusion. The prominent NSL-KDD dataset was utilized for training and testing in this study. The accuracy obtained for two and five classes was 84.7 percent and 90.5 percent, respectively. Additionally, the identification speed was 15 µs and 14 µs, respectively, indicating that real-time detection is feasible. |
---|