AI and IoT Powered Intelligent System for Real Time Elephant Movement Detection and Prevention of Human-Elephant Conflict (HEC) in Chhattisgarh State-1.pptx

AI and IoT Powered Intelligent System for Real Time Elephant Movement Detection and Prevention of Human-Elephant Conflict (HEC) in Chhattisgarh State

1. Intelligent Flow Rule Optimization in SDN Team: Chandana, Noor Hafsa, Syed Yousuf, Chiranth S Guide: Ms. Shambhavi K A Assistant Professor Dept. of Cyber Security ATME College of Engineering Mysore

2. Abstract Introduction SDN separates control and data planes. Switches use TCAM memory (fast but costly, limited). Short-lived or malicious flows exhaust TCAM . Project: ML-based controller to filter & optimize rules. Results: Less TCAM use, high accuracy, better performance. SDN enables centralized, programmable networking. TCAM stores flow rules but is a scalability bottleneck . TCAM exhaustion → packet drops, latency, poor throughput. Goal: Smart ML-driven flow rule management.

3. 04 Validate with Simulations Test on Mininet with attacks 03 Optimize TCAM Usage Save memory without losing speed 02 Compare Classifiers DT, RF, SVM, DNN for accuracy 01 Design ML Module Real-time flow classification system Objectives

4. Problem Statement All flows installed in SDN Hard to distinguish flow types Network slowdown Switches store everything → TCAM overload Normal, short-lived, and malicious traffic Critical performance drop & vulnerability

5. Existing System Step 1 Static Rule Aggregation 01 Step 2 Timeout Eviction 02 Step 3 Heuristic Prioritization 03 Step 4 Batch Rule Updates 04 Step 5 Data-driven approach The Earth is the third planet from the Sun 05 Combines rules into fewer entries. Problem: loses precision, may allow malicious traffic. Removes rules after fixed time. Problem: may drop good flows or keep junk too long. Prioritizes flows based on simple metrics. Problem: often misclassifies important or attack flows. Installs all flows in bulk. Problem: wastes TCAM, slows performance.

6. Proposed System 01 Machine Learning for Smarter Flow Control Analyzes traffic features to optimize rules and block malicious flows early. 02 Features packet count, time, duration, protocol 03 Predict essential vs. discardable flows 04 Install only important rules Drop junk/malicious traffic early

7. Literature Survey ML reduced rules by 40–65% 01 Hybrid proactive-reactive up to 30% lower latency 03 Adaptive routing strategies → improved throughput by 20–35% 05 TCAM savings: up to 55% 02 Cost-aware eviction → balanced delay & memory 04 Energy-efficient approaches → reduced power consumption in network devices 06

8. Software Requirements OS Ubuntu 20.04 / CentOS 8 01 Tools Jupyter , Wireshark, Git, Docker 05 Controller Ryu (OpenFlow 1.3) 03 Wireshark 3.6+ for packet-level verification 07 Emulator Mininet 2.3+ 02 Data Streaming Apache Kafka 2.8+ 06 ML Python 3.8+, scikit-learn, TensorFlow 04 Version Control Git 2.30+ Docker 20.10+ 08

9. Hardware Requirements 01 CPU Intel i5 / Ryzen7+ RAM 32 GB (64 GB recommended) 02 GPU NVIDIA RTX 3060 Ti (8 GB VRAM+) 03 Virtualization Intel VT-x / AMD-V 05 Network Gigabit Ethernet 04 Controller Ryu Controller v4.34 06

10. References [1] https://www.nature.com/articles/s41598-024-65721-x [2] https://doi.org/10.1109/COMST.2021.310456 [3] https://doi.org/10.1016/j.jksuci.2018.01.005 [4] https://doi.org/10.1007/978-3-030-50029-0_15 [5] https://www.sciencedirect.com/science/article/pii/S1389128619305678