• FARHAN
• MD HUZAIF SHAH
• ADITYA SHIVAPUJE
• ARCHANA SATHELLI

The proliferation of artificial intelligence (AI) has led to significant advancements in biometric security and digital content verification, but it has also enabled sophisticated threats such as spoofing attacks and deepfake manipulations. The Secure Vision project addresses these dual challenges by integrating MobileNet and ResNeXt models into a unified system. MobileNet is employed for real-time anti-spoofing detection, focusing on liveness cues like eye blinks and subtle facial movements, while ResNeXt specializes in identifying deepfake-specific artifacts, including texture inconsistencies and lighting anomalies. By training on diverse datasets such as ASVspoof, FaceForensics++, and the Deepfake Detection Challenge (DFDC), the system achieves robust performance and generalizability. The integrated architecture delivers high accuracy (97.8%) and real-time processing capabilities (50ms per frame), making it suitable for applications in biometric access control, online identity verification, and media authenticity validation. This paper discusses the methodology, performance metrics, and potential for future enhancements, including multi-modal integration and continuous learning frameworks, to ensure the system evolves alongside emerging threats.

MobileNet, ResNeXt, Real-Time Processing, FaceForensics++ Dataset, ASVspoof Dataset

"SECURE VISION INTEGRATED ANTI-SPOOFING AND DEEPFAKE DETECTION SYSTEM USING MOBILENET AND RESNEXT", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.11, Issue 12, page no.e773-e777, December-2024, Available :http://www.jetir.org/papers/JETIR2412489.pdf

"SECURE VISION INTEGRATED ANTI-SPOOFING AND DEEPFAKE DETECTION SYSTEM USING MOBILENET AND RESNEXT", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.11, Issue 12, page no. ppe773-e777, December-2024, Available at : http://www.jetir.org/papers/JETIR2412489.pdf