Network Intrusion Detection in Simulated Military Environment Using Machine Learning and Blockchain

Network Intrusion Detection Systems are critical in bolstering an organization's security infrastructure. The goal of our paper is to design and build a specialized intrusion detection system for simulated military scenarios. The solution classifies network accesses using machine learning techniques and seamlessly integrates the Ethereum blockchain for secure anomalous access record-keeping and auditing. The machine learning component classifies the records as normal or anomalous using the ensemble learning technique, which is the random forest classifier. Feature selection techniques like Variance Threshold, Spearman's correlation coefficient, Pearson's correlation coefficient, Mutual Information and Chi square tests were used to improve the accuracy of the model and reduce dimensionality. The final accuracy of the model was brought to 96.69%. It then sends the records to the blockchain environment for storage. To ensure data integrity and transparency, anomalous access events will be time-stamped, hashed using IPFS, and published to the blockchain. By providing a hash value and a unique id as input, the user will be able to add new anomalous entries. Furthermore, the user will be able to examine the hashed anomalous data for each csv file using the proper id.