MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641092285 A) filed by Dr. Safia Naveed. S; and Krithikaa Venkat Vs on July 30, 2026, for Revealing The Truth : A Groundbreaking Deep Learning Strategy For Detecting Deep Fake Audio.
Inventors include Dr. Safia Naveed. S; and Krithikaa Venkat Vs.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: The rapid advancement of deep learning technologies has led to the emergence of highly realistic synthetic_audio,_commonly referred to as deepfake audio.-VVhile- such innovations have beneficial applications in entertainment, virtual assistants, and accessibility, they also pose significant threats in areas such as cybersecurity, financial fraud, misinformation, and digital forensics. Detecting deepfake audio has therefore become a critical challenge due to the increasing sophistication of voice synthesis and cloning techniques. This invention presents a novel deep learning-based framework for the effective detection of deepfake audio. The proposed system integrates a multi-stage pipeline consisting of audio preprocessing, feature extraction, and classification using advanced neural network models. Initially, the input audio undergoes preprocessing to remove noise and normalize the signal. Subsequently, discriminative features such as spectral and temporal characteristics are extracted to capture subtle differences between genuine and manipulated audio signals. These features are then fed into a deep learning model trained on benchmark datasets such as ASVSpoof to ensure robustness and generalization. The system is designed to achieve high accuracy while addressing challenges such as environmental noise, variability in speech patterns, and evolving spoofing techniques.
Disclaimer: Curated by HT Syndication.
