A Bio-Inspired Path to Fake Profile Detection: Revisiting Linear Models through Grasshopper Optimization

Abstract

Social media platforms continue to struggle with the proliferation of fake user profiles, which undermine trust and facilitate the spread of misinformation. While deep learning and complex ensemble models dominate recent solutions, this study revisits the power of simpler classifiers when paired with intelligent feature selection. We introduce a hybrid framework that couples the Grasshopper Optimization Algorithm with Logistic Regression for detecting fake profiles in online social networks. The proposed bio-inspired algorithm mimics the collective foraging behavior of grasshoppers to optimize the feature space, resulting in a compact and highly discriminative set of inputs. Evaluated on different datasets from social media platforms, the proposed model not only outperforms six standard classifiers but also challenges the assumption that fake profile detection requires non-linear modeling. Achieving impressive accuracy and strong F1 performance, our approach shows that nature-inspired metaheuristics can elevate classical models to competitive levels. These results suggest a promising direction for lightweight, interpretable, and scalable solutions in social cybersecurity.