Across various areas, optimal foraging algorithms (OFA) have demonstrated remarkable efficacy in various machine learning applications. Due to its graph representation structure, applying the knowledge learned to textual material is highly difficult. This study addresses safe wireless systems and mobile computing applications by utilizing novel graph architectures and protection strategies. Using OFA and support vector machines (SVM), we create an intrusion detection system (IDS) that can detect adversarial inversion attempts within a network system. It uses enemies that are both abnormal and normal. It develops attack signatures, generating signature continuously, and updates the IDS signature repository. In conclusion, the efficacy and efficiency of the suggested framework with Random Forest are assessed using the assessment indicators, which include latency rates and throughput. In terms of latency rate and throughput, the suggested model (SVM with OFA) based on adversarial inversion attacks performed better and more efficiently than conventional models, with detection rates of 93.68% and 95.32%, respectively. Additionally, accuracy comparisons of the feature dataset with 90.4% and 91%, respectively, and accuracy comparison of the suggested system (SVM with OFA) with the other classifier are shown in this article.