Improvement of Packet Delivery Ratio in MANET Using ADLR: A Modified Regularization-Based Lasso Regression

Abstract

Mobile Ad hoc Networks (MANETs) are characterized by their dynamic topology, limited bandwidth, and unpredictable link conditions, which make reliable packet delivery a significant challenge. Regression techniques have been previously utilized in Mobile Ad hoc Networks (MANET) with a high success rate. However, a modification of Lasso regression is still required to effectively mitigate attacks while improving service quality. In our research paper, we developed an innovative method that integrates a modified Lasso regression analysis with a regularization term based on absolute deviation. By comparing the modified Lasso regression algorithm with other regression techniques, such as linear regression, ridge regression, Least Absolute Deviation (LAD) regression, Method of Least Square (MLS) regression, and Ridge regression, we demonstrate its superior performance over the others. During the Wormhole attack, we attempted to enhance the packet delivery ratio, reduce delay variance, etc., using our algorithm. It was discovered that the Absolute Deviated Lasso Regression (ADLR) algorithm outperformed others significantly when dealing with large datasets. Through experimental evaluation, we have demonstrated that our modified algorithm significantly reduces errors and maximizes the packet delivery fraction by more than 90% compared to other regression techniques. The highest packet delivery ratio reaches 92%, while the lowest packet loss ratio is 24% when using our ADLR algorithm.