Abstract
Fake profiles, on Instagram, pose significant challenges in online spaces. These profiles are fabricated with the intention to mislead and deceive by assuming the identity of genuine persons or institutions. These individuals often disseminate false information, participate in spamming, or carry out criminal acts, which negatively impact user confidence and online safety. Detecting these fake profiles involves analysis of various factors like account activity, behavior patterns, content, and network interactions to distinguish these from genuine users. Advancements in machine learning and AI techniques play a crucial role in developing robust detection models to mitigate the impact of these deceptive profiles. This paper delves into the intricate realm of fake profile detection on social-media platforms. The proposed model exhibits commendable performance metrics, showcasing an accuracy of 83.84%. Delving into specifics, for the identification of fake profiles, precision stands at 80.65%, recall at 80.16%, and an F1-Score of 80.41%. When detecting real profiles, precision, recall, and F1-Score excel significantly higher at 86.06%, 86.43%, and 86.24% respectively. These impressive metrics are achieved through a sophisticated approach leveraging DistilBERT for text processing, SMOTE for handling imbalanced data, and Random Forest for classification to detect fake profiles on Instagram based on user biography length. In essence, this research contributes substantial insights and advanced methodologies to the intricate domain of fake profile detection. It aspires to fortify the trustworthiness and credibility of online identity verification systems, serving as a foundational stone for future advancements in the field.