Estimation of Optimal Time-to-Live Values for Mitigating Router Spoofing Attacks Utilizing Lasso and Ridge Regression Machine Learning Techniques

Router spoofing attacks pose a significant threat to network security, making accurate estimation of Time-to-Live (TTL) values in network traffic essential for robust defense. In our study, Lasso and Ridge regression techniques were used for TTL values prediction. Lasso regression has reduced model complexity by eliminating non-essential features and Ridge regression worked fairly well in situations with multiple linear relationships. Our models trained with real network traffic data have high accuracy in determining the better TTL settings, and as a result making defence against router spoofing attack efficient. The analysis results underscored the performance of both regression models in identifying forged packets. Fortunately, lasso regression achieved a notable accuracy rate of $\mathbf{9 6. 4 \%}$, with 135 true positives and only 5 false positives out of 140 predictions. Ridge regression showed an accuracy of $\mathbf{9 2. 8 \%}$, with 130 true positives and 10 false positives. These findings indicate that Lasso regression is particularly effective for TTL prediction and forged packet detection, especially in noisy environments.