![AutoML-Based Framework for
AutoML-Based Framework for
Optimization of Intrusion Detection
Optimization of Intrusion Detection
Pipelines in Network Security
Pipelines in Network Security
Author :
R.Kiruba buri
[ Department of CSE, University College of Engineering, Anna University,
Pattukkottai Campus, Tamil Nadu-614701 ]*
Co-authors:
2. K. Swaminathan
[Department of ECE, University College of Engineering, Rajamadam,
Pattukkottai]
3. S.Sundarsingh
[Department of ECE, University College of Engineering , Thirukkuvalai]
4. K.Sankar
[Department of EEE, University College of Engineering , Pattukottai.]
5. Yuvarajan P
[Department of EEE, University College of Engineering , Panruti]](https://image.slidesharecdn.com/332iccct2025-250325104227-a9d7ad24/85/AutoML-Based-Framework-for-Optimization-of-Intrusion-Detection-Pipelines-in-Network-Security-2-320.jpg)
![Literature Survey
No.
Methodology
Name
Advantage Key Findings
[1] SVM-Based IDS
High accuracy for known
attacks
Achieved 90% accuracy
but struggled with new
attacks
[2] Deep Learning IDS Improved feature extraction
CNN-based IDS
increased detection
rates
[3] XGBoost IDS Fast and accurate
XGBoost outperformed
traditional classifiers
[4]
Federated Learning
IDS
Better security and
scalability
Federated approach
protected user privacy
[5] AutoML IDS
Fully automated model
selection
Reduced manual
intervention and tuning
time](https://image.slidesharecdn.com/332iccct2025-250325104227-a9d7ad24/85/AutoML-Based-Framework-for-Optimization-of-Intrusion-Detection-Pipelines-in-Network-Security-6-320.jpg)
More Related Content
Similar to AutoML-Based Framework for Optimization of Intrusion Detection Pipelines in Network Security (20)
Recently uploaded (20)
AutoML-Based Framework for Optimization of Intrusion Detection Pipelines in Network Security
- 1. 7th International Conference on Communication and Computational Technologies (ICCCT2025) For Paper ID: 332 Paper Title: AutoML-Based Framework for Optimization of Intrusion Detection Pipelines in Network Security 14 feb2025
- 2. AutoML-Based Framework for AutoML-Based Framework for Optimization of Intrusion Detection Optimization of Intrusion Detection Pipelines in Network Security Pipelines in Network Security Author : R.Kiruba buri [ Department of CSE, University College of Engineering, Anna University, Pattukkottai Campus, Tamil Nadu-614701 ]* Co-authors: 2. K. Swaminathan [Department of ECE, University College of Engineering, Rajamadam, Pattukkottai] 3. S.Sundarsingh [Department of ECE, University College of Engineering , Thirukkuvalai] 4. K.Sankar [Department of EEE, University College of Engineering , Pattukottai.] 5. Yuvarajan P [Department of EEE, University College of Engineering , Panruti]
- 3. Overview of Intrusion Detection Optimization with AutoML AutoML automates model selection and tuning for IDS. XGBoost is used for better classification accuracy. The framework reduces human intervention and improves efficiency. Adapts to evolving network security threats. Enhances scalability in modern network environments.
- 4. Research Problem Challenges in Intrusion Detection Systems Traditional IDS require manual feature selection and tuning. High False Positive Rate (FPR) reduces reliability. Difficulty in handling large-scale datasets efficiently. Struggles to adapt to evolving cyber threats
- 5. Objectives Research Goals and Contributions Develop an AutoML-based framework for IDS. Optimize model selection and hyper- parameter tuning. Enhance detection accuracy and reduce False Positive Rate (FPR). Improve scalability and real-time adaptability
- 6. Literature Survey No. Methodology Name Advantage Key Findings [1] SVM-Based IDS High accuracy for known attacks Achieved 90% accuracy but struggled with new attacks [2] Deep Learning IDS Improved feature extraction CNN-based IDS increased detection rates [3] XGBoost IDS Fast and accurate XGBoost outperformed traditional classifiers [4] Federated Learning IDS Better security and scalability Federated approach protected user privacy [5] AutoML IDS Fully automated model selection Reduced manual intervention and tuning time
- 7. Proposed Methodology: Proposed Methodology: AutoML-Based Intrusion AutoML-Based Intrusion Detection Pipeline Detection Pipeline Data Preprocessing (Feature selection & normalization). Model Selection (AutoML scans for the best classifiers). Hyperparameter Tuning (Optimized for accuracy and FPR). Evaluation Metrics (Accuracy, Precision, FPR, and Detection Time).
- 8. Traditional IDS: Manual feature selection, rule-based detection. Machine Learning-Based IDS: Automated pattern recognition but requires hyperparameter tuning. AutoML-Based IDS: Automates model selection, reduces human intervention, and adapts dynamically to threats. XGBoost in IDS: High classification accuracy, scalable for large datasets. Comparison of IDS Approaches
- 9. Automated Model Selection: Reduces human effort in choosing the best model. Improved Detection Accuracy: High precision in identifying intrusions. Scalability: Handles large-scale network traffic efficiently. Low False Positive Rate (FPR): Minimizes incorrect threat alerts. Real-Time Adaptability: Adjusts to evolving security threats dynamically Advantages of the Proposed AutoML Framework
- 10. Challenges in AutoML-Based IDS Challenges in AutoML-Based IDS Computational Complexity: Requires high processing power for model training. Potential Overfitting: Risk of models adapting too much to training data. Adversarial Attacks: Attackers may manipulate input data to evade detection. Integration with Legacy Systems: Compatibility issues with older network architectures. Data Privacy Concerns: Handling sensitive network traffic securely.
- 11. Detection Accuracy Analysis (Figure 2) Detection Accuracy Analysis (Figure 2) Accuracy range: 0.92 to 0.97. Threshold set at 0.95. Iterations 1, 2, and 8 fall below threshold. Iterations 3 to 7 exceed threshold (0.96 to 0.97). Indicates model effectiveness but requires optimization for lower-performing iterations.
- 12. False Positive Rate (FPR) Analysis (Figure 3) False Positive Rate (FPR) Analysis (Figure 3) FPR range: 0.01 to 0.05. Threshold: 0.05. Iterations 3, 4, and 8 exceed threshold. Iterations 1, 2, 5, 6, and 7 remain below threshold. Indicates reliable detection with minimal false positives.
- 13. Precision Analysis (Figure 4) Precision Analysis (Figure 4) Precision range: 0.87 to 0.96. Threshold: 0.90. Iterations 1, 2, and 8 fall below threshold. Iterations 3 to 7 exceed threshold. Highlights the model's ability to correctly identify intrusions.
- 14. Detection Time Analysis (Figure 5) Detection Time Analysis (Figure 5) Detection time range: 1.5 to 3.0 sec. Threshold: 2.0 sec. Iterations 3, 4, and 8 exceed threshold. Iterations 1, 2, 5, 6, and 7 remain within acceptable limits. Indicates potential delays in real-time detection.
- 15. Conclusion AutoML-based IDS enhances detection accuracy and scalability. XGBoost ensures high precision in identifying threats. FPR remains low, improving reliability. Challenges include computational complexity and adversarial attacks. Future work: Optimize model efficiency and integrate with advanced security frameworks.
- 16. Future Work Future enchantment will focus on improving scalability by exploring distributed and federated learning mechanism, allowing the system to operate various multiple network nodal points, thus minimizing the computational load on central server infrastructure. Furthermore, incorporating hybrid system that combine anomaly determining with signature -depending mechanism could improve the determining of zero-day attacks and minimize false positives.
- 17. References : References : El Rajab, M., Yang, L., &Shami, A. (2024). Enhancing Network Intrusion Detection: An AutoML Pipeline with Efficient Digital Twin Synchronization. Authorea Preprints. Glavan, A. F., &Croitoru, V. (2023, June). Autoencoders and AutoML for intrusion detection. In 2023 15th International Conference on Electronics, Computers and Artificial Intelligence (ECAI) (pp. 1-4). IEEE. Papanikolaou, A., Alevizopoulos, A., Ilioudis, C., Demertzis, K., &Rantos, K. (2023). An autoML network traffic analyzer for cyber threat detection. International Journal of Information Security, 22(5), 1511-1530. Schubert, D., Eikerling, H., &Holtmann, J. (2021). Application-Aware Intrusion Detection: A Systematic Literature Review, Implications for Automotive Systems, and Applicability of AutoML. Frontiers in Computer Science, 3, 567873. Sezgin, A., &Boyac脹, A. (2023). AID4I: An Intrusion Detection Framework for Industrial Internet of Things Using Automated Machine Learning. Computers, Materials & Continua, 76(2). Kalyanaraman, K., & Prabakar, T. N. (2024). Enhancing Women's Safety in Smart Transportation Through Human-Inspired Drone-Powered Machine Vision Security. In AI Tools and Applications for Womens Safety (pp. 150-166). IGI Global. Liuliakov, A., Hermes, L., & Hammer, B. (2023). AutoML technologies for the identification of sparse classification and outlier detection models. Applied Soft Computing, 133, 109942. Yang, L., El Rajab, M., Shami, A., &Muhaidat, S. (2024). Enabling AutoML for Zero-Touch Network Security: Use-Case Driven Analysis. IEEE Transactions on Network and Service Management. Filippou, K., Aifantis, G., Papakostas, G. A., &Tsekouras, G. E. (2023). Structure learning and hyperparameter optimization using an automated machine learning (AutoML) pipeline. Information, 14(4), 232.
- 18. Thank You