PDF Malware Detection Toward Machine Learning Modeling With Explainability Analysis

The project "A Hybrid Deep Learning Framework for Deep Fake Detection Using Temporal and Spatial Features" is designed to address the growing concern of deep fake content across three primary media types: images, videos, and audio. For image-based deep fake detection, the system utilizes advanced computer vision models such as YOLOv8, YOLOv10, Fast-RCNN, and EfficientDet. These models are specifically tailored to detect fake facial images by analyzing spatial features. For video content, InceptionNet is combined with a GRU (Gated Recurrent Unit) network, enabling the model to identify deep fakes through both temporal and spatial feature extraction. Additionally, a CNN-based model is employed for detecting deep fake audio, capturing anomalies in sound patterns. The project also includes the development of a web interface using HTML, CSS, and Flask, providing an intuitive platform for users to upload images, videos, and audio for analysis. Users can log in, register, and easily submit files for deep fake detection, receiving predictions based on the uploaded media. This system contributes to enhancing digital media security by providing an effective solution for deep fake identification.

Keywords:

Deep Fake Detection, Hybrid Deep Learning, Temporal Features, Spatial Features, YOLOv8, YOLOv10, Fast-RCNN, EfficientDet, InceptionNet, GRU, CNN, Audio Detection, Image Detection, Video Detection, Flask Web Interface, Digital Media Security.