ForensicFlow: A Tri-Modal Adaptive Network for Robust Deepfake Detection

Mohammad Romani · 2025

Read ForensicFlow: A Tri-Modal Adaptive Network for Robust Deepfake Detection by Mohammad Romani (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-14554.txt.

Description

Mohammad Romani presents ForensicFlow, a deepfake detector that fuses three kinds of forensic evidence. Modern fakes hide flaws that a single network branch can miss. One branch checks global visual inconsistencies with a compact convolutional model. A second branch hunts fine texture anomalies with a transformer. A third branch reads spectral noise patterns.

Attention based pooling weights frames with the strongest evidence most heavily. Adaptive fusion then weights each branch by forgery type. Trained on a celebrity deepfake data set, the model reports high accuracy and area under the curve. Ablation tests confirm each branch helps. Heat maps show focus on true manipulation zones like face edges.

Catalog record

CREATORMohammad Romani
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE26 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-14554

Browse the full library · About the Tower of Babel