Content Retrieval in Information Centric Networks Using Machine Learning Models
B. Surya Samantha, Nitul Dutta, Krishna Delavadia · 2024
In Information Centric Network (ICN) consumer generates an interest packet to retrieve data. The intermediate nodes forwards interest packet to reach content location. The ICN does not use any intelligence in forwarding interest. In this paper, an intelligent forwarding scheme is proposed based on Machine Learning (ML). For that, a routing and forwarding dataset derived from simulated ICN topology is used. In the first stage, a Content Routers (CRs) are trained using ML models. Then, CRs use the acquired knowledge to forward future interest packets. It is found that the use of ML significantly improves the performance in terms of retrieval time and network throughput. The said dataset is created from a simulation scenario designed in ndnSim-2.0. Eighteen models are used in predicting the content location. The experimental evaluation reveals that the Extra Tree (ET), Random Forest (RF), and Decision Tree (DT) perform with 98% accuracy. The results of the machine model were verified by creating a random ICN topology and examining the success rate for ML-guided interest forwarding. A comparison is made with forwarding techniques with and without using ML techniques. In comparison to non-ML variations, ML-guided forwarding techniques perform better in content retrieval with nearly 20% improvement in performance.