Dynamic Algorithmic Configuration for Enhanced Malware Detection

As edge-based decisions continue to grow in demand, organizations are increasingly seeking user data, raising privacy concerns. Federated Learning (FL) addresses this problem but is typically dependent on computationally intensive Neural Network (NN) models. This work introduces a novel lightweight approach by integrating Support Vector Machines (SVMs) into FL frameworks. Optimized methodology SVM with the compatibility of SGD is used where the meta-trained controller dynamically adapts the learning rate to the improvement of models. The proposed method shows equal performance to those based on the NN-based FL methods but drastically reduces the cost of computation for being applicable to resource-constrained environments.