Utilizing blockchain-secured deep packet inspection for trusted differentiated services

Abstract

The recent developments in networking have led to the generation of massive volumes of network traffic, creating many obstacles in network traffic management. The traditional deep packet inspection (DPI) is used to inspect, identify, classify and manage Internet Protocol (IP) based network packets from massive volume of traffic flow within the connected network. However, with the integration of DPI at various inspection points at the network edges, it creates certain challenges for the connected networks. Also, there are security challenges posed with conventional deep packet inspection related to the identification of tampered, manipulated and exploited IP network packets within the network transmission classified for critical data having higher priority access using differentiated service code point (DSCP). The purpose of this research is to design and develop a blockchain-secured deep packet inspection system to secure the DSCP tagging and examine the impact of blockchain on DPI system as well as on real-time operations. Moreover, the study explores the use of deep learning models due to its efficient learning hierarchical representation of features through series of layers as well as with memory for its ability to extract features through time series data. Also, the use of blockchain in the spectrum of DPI to enhance the performance and security of network. The study has utilized four deep learning neural classifiers including Autoencoders, Convolutional Neural Network, Long-Short Term Memory and Recurrent Neural Network having accuracies of 92.03%, 96.88%, 99.64% and 99.72% respectively. It was discovered the RNN performed well around 99.72% than the rest of the adopted classifiers. There is an ensemble approach used based on soft-voting mechanism with an accuracy of 99.28% having all the features of deployed models into one single model. Additionally, the ethereum-based blockchain is used to develop and deploy blockchain into DPI ecosystems, storing critical IP packets information moving forward from the classification phase of the neural network. The IP packets data was also stored in NoSQL database systems for future analysis on different use-cases. The objective of this study contributes to the analysis of the security issues related to DSCP faced by DPI with the growing volume of network data.