Related Experiment Videos
DeepShieldIDS: an AI-powered intrusion detection system leveraging HybridIDSNet for robust network security
Sayyada Mubeen1, Balakrishna Gudla2
1Department of computer Science and Engineering, Malla Reddy university, Hyderabad, Telangana, India. 2232cs010018@mallareddyuniversity.ac.in.
Scientific Reports
|May 20, 2026
Summary
DeepShieldIDS enhances network intrusion detection using a hybrid deep learning model. This framework effectively captures temporal network traffic patterns, significantly improving threat detection accuracy and generalization capabilities.
Area of Science:
- Cybersecurity
- Artificial Intelligence
- Network Security
Background:
- Traditional Intrusion Detection Systems (IDSs) struggle with modern cyber threats due to static feature representations that fail to capture temporal network traffic dynamics.
- Existing deep learning methods face challenges in feature diversity, class imbalance, and cross-environment generalization for network intrusion detection.
Purpose of the Study:
- To introduce DeepShieldIDS, a novel hybrid deep learning framework designed to overcome the limitations of conventional network intrusion detection systems.
- To improve network traffic classification accuracy and generalization by integrating spatial, temporal, and attention-based feature learning.
Main Methods:
- Developed DeepShieldIDS, a hybrid framework combining Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and an attention mechanism.
- Utilized spatial feature extraction, temporal sequence modeling, and feature-weighting attention within a unified pipeline for traffic classification.
- Conducted experiments on CIC-IDS2017 and UNSW-NB15 datasets in independent and cross-dataset evaluation settings.
Main Results:
- Achieved high detection accuracies of 97.98% and 98.02% with ROC-AUC scores of 99.05% and 99.12% on benchmark datasets.
- Demonstrated superior performance compared to classical machine learning models (Random Forest, SVM, XGBoost) and baseline deep learning models (Categorical CNN, LSTM).
- Cross-dataset validation indicated improved generalization capabilities when handling heterogeneous traffic distributions.
Conclusions:
- The proposed HybridIDSNet framework offers a scalable solution for network intrusion detection, effectively addressing limitations in feature representation and generalization.
- DeepShieldIDS provides a deployment-oriented architecture and a formalized pre-processing procedure for practical application in network security.
- Hybrid deep learning approaches integrating CNNs, LSTMs, and attention mechanisms represent a significant advancement in detecting evolving cyber threats.