Deep Fake Detection with Hybrid Activation Function Enabled Adaptive Milvus Optimization-Based Deep Convolutional Neural Network
Harish Mashetty, Naresh Erukulla, Sandeep Belidhe, Naresh Jella, Varun reddy Pishati, Bharath Kumar Enesheti · 2025
Deep fake detection plays an essential role in digital world platforms that effectively identify the fake content of audio and video streams, which disseminated more on social media platforms in recent years. Several researchers have undertaken to detect the fake context, which faced certain challenges such as information authentication, media security, stolen identities, damaging someone's personal life, and so on. Thus, to enhance the detection accuracy and to overcome the challenges of existing methods, the Hybrid activation function enabled adaptive Milvus optimization-based Deep Convolutional Neural Network (HA2MilO-DCN) model is developed in the research. The integration of the hybrid activation function with the base DCNN model effectively enhanced the scalability and efficiency to achieve the desired detection results. Furthermore, the adaptive Milvus optimization (AMilO) effectively tunes the hyperparameter of the model, which is inspired by the social behavior of Milvus and aids in attaining prominent detection outcomes. Therefore, the HA2MilO-DCN attains a high range of detection accuracy with performance metrics such as Accuracy, Recall, and Precision. Under the utilization of the Faceforensics++ dataset, the model attains effective detection outcomes when compared with other conventional methods. The obtained evaluation metrics values are 95.72%, 94.91%, and 96.52% respectively.