Anita_Asienga_Dissertation_Presentation.pptx

Smart grids use IoT to improve energy efficiency, but this also increases exposure to cyber threats like DDoS attacks. Traditional defenses can’t handle the dynamic nature of these attacks. This study develops an LSTM-based intrusion detection model to identify and mitigate DDoS attacks in smart grids.

Using open datasets such as CIC-DDoS2019, the research follows the CRISP-DM framework — from data preparation to model evaluation. The LSTM model learns traffic patterns and detects abnormal flows linked to DDoS activity. Results show that LSTM achieves higher accuracy and reliability compared to other ML models.

The study fills a research gap by focusing on real-time, adaptive DDoS detection for smart grids. It contributes to building more resilient and secure energy systems and sets a foundation for future work involving live smart grid environments and hybrid detection models.

1. Introduction Smart grids, powered by IoT, revolutionize energy management but increase cyber risk. DDoS attacks disrupt grid communication and operations. Objective: Build an LSTM-based IDS to detect and mitigate DDoS in smart grids.

2. Problem Statement Integration of IoT with Operational Technology expands attack surface. Traditional defenses fail against adaptive DDoS attacks. Need for intelligent, dynamic ML-based intrusion detection.

3. Research Objectives General: Develop an ML-based DDoS detection model for smart grids. Specific Objectives: 1. Analyze effects of DDoS on smart grids. 2. Assess existing mitigation strategies. 3. Design and test an LSTM-based model. 4. Validate using benchmark datasets.

4. Research Questions 1. What are the effects of DDoS attacks on smart grids? 2. How do current defenses mitigate them? 3. How can an LSTM model detect and classify attacks? 4. How effective is it when tested on real data?

5. Scope and Limitations Scope: - Focus on ML-based detection using open datasets. - Emphasis on modeling, not real deployment. Limitations: - Relies on simulated datasets. - Model performance may vary in real environments.

6. Research Justification Smart grids are critical infrastructure vulnerable to DDoS. Traditional static defenses are inadequate. LSTM models detect temporal patterns and anomalies in traffic. Bridges research gap and informs policy and industry resilience.

7. Literature Review I Technological & Theoretical Review: - Smart grid architecture & vulnerabilities. - DDoS types: volumetric, protocol, application-layer. - Theories: Statistical Learning, Information, Game Theory.

8. Literature Review II Empirical & Research Gaps: - ML approaches: ANN, SVM, hybrid, blockchain. - Limited focus on smart grids. - Need for dynamic ML models tailored to energy networks.

9. Methodology I: Design & Data Experimental research design. Data: CIC-DDoS2019 & CSE-CIC-IDS2018 datasets. Sampling: Stratified, 70% train, 20% validation, 10% test.

10. Methodology II: Model Development CRISP-DM Framework: - Business Understanding - Data Understanding - Data Preparation - Modelling with LSTM - Evaluation & Deployment

11. Methodology III: Evaluation & Ethics Metrics: Accuracy, Precision, Recall, F1-Score. Ethical Focus: - Data privacy (open-source datasets). - Bias mitigation in training. - Transparency in reporting.

12. Conclusion LSTM-based IDS enhances smart grid resilience against DDoS. Offers adaptive, real-time defense approach. Future work: Real-world implementation and multi-model fusion.