MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641069913 A) filed by Sasi Institute Of Technology & Engineering; A. V. S. Siva Rama Rao; M Satya Srinivas; Sivaramarajasasi. Ac. In; and V Srinivas on June 04, 2026, for Hybrid Deep Learning And Machine Learning Approach For Detecting Deepfake Content.

Inventors include Sasi Institute Of Technology & Engineering; A. V. S. Siva Rama Rao; M Satya Srinivas; Sivaramarajasasi. Ac. In; and V Srinivas.

The application for the patent was published on June 12, 2026, under issue no. 24/2026.

Abstract: The present invention relates to a system and method for detecting deepfake images using a hybrid deep learning and machine learning approach. With the rapid advancement of artificial intelligence technologies, the creation of highly realistic manipulated images has increased, posing serious threats to digital authenticity, privacy, and information reliability. The proposed system integrates multiple convolutional neural network architectures, including DenseNet121, ResNet18, EfficientNet-B4, and a custom-designed Random CNN model, to extract diverse and discriminative visual features from input images. Each model independently analyzes the image and produces prediction probabilities, which are then combined using ensemble learning techniques such as probability averaging and majority voting to generate a final classification. The system incorporates preprocessing steps such as image resizing, normalization, and tensor conversion to ensure compatibility with deep learning models. It is trained on a large and diverse dataset of approximately 30,000 real and deepfake images, enabling improved generalization and robustness. The invention further includes a web-based interface that allows users to upload images and receive real-time detection results along with confidence scores. Experimental results demonstrate that the proposed hybrid approach achieves high detection accuracy in the range of 96–98%. The invention provides an efficient, scalable, and reliable solution for automated deepfake detection, with applications in digital media verification, cybersecurity, and fraud prevention.

Disclaimer: Curated by HT Syndication.