HYBRID REALNESS VERIFICATION SYSTEM

With the rapid advancement of artificial intelligence, the creation of synthetic media, commonly known as deepfakes, has become increasingly realistic and accessible. This growing sophistication poses significant threats to digital security, media authenticity, and online trust. To counter these challenges, this work presents a Hybrid Realness Verification System (HRVS) a multimodal deepfake detection framework designed to assess the authenticity of image, video, and audio data through domain-specific machine learning techniques.For visual analysis, HRVS integrates a vision transformer-based model to identify subtle inconsistencies such as facial distortion, unnatural texture patterns, and absence of micro-expressions. Audio verification is achieved by transforming voice signals into Mel spectrograms and evaluating them using a convolutional neural network trained to detect artifacts of AI-generated speech, including irregular pitch, rhythm, and tonal energy.The individual predictions from these modules are combined through a fuzzy logic-based decision engine, enabling the system to manage uncertain or borderline cases with adaptive reasoning instead of fixed thresholds. The overall architecture is implemented on a Flask backend, supporting real-time user interaction, scalability, and seamless integration.By fusing multimodal feature analysis with intelligent decision-making, HRVS enhances detection reliability and interpretability. Its applications span security verification, content moderation, digital forensics, and identity authentication, contributing to a safer and more trustworthy digital ecosystem.