ProtoExplorer: Interpretable Forensic Analysis of Deepfake Videos using Prototype Exploration and Refinement

View on arXiv ← Back to list

Authors: Merel de Leeuw den Bouter, Javier Lloret Pardo, Zeno Geradts, Marcel Worring

Published: 2023-09-20 09:03:56+00:00

AI Summary

ProtoExplorer is a visual analytics system for exploring and refining prototype-based deepfake video detection models. It allows forensic experts to interactively improve model interpretability and reduce bias while maintaining detection accuracy by visualizing and manipulating spatio-temporal prototypes.

Abstract

In high-stakes settings, Machine Learning models that can provide predictions that are interpretable for humans are crucial. This is even more true with the advent of complex deep learning based models with a huge number of tunable parameters. Recently, prototype-based methods have emerged as a promising approach to make deep learning interpretable. We particularly focus on the analysis of deepfake videos in a forensics context. Although prototype-based methods have been introduced for the detection of deepfake videos, their use in real-world scenarios still presents major challenges, in that prototypes tend to be overly similar and interpretability varies between prototypes. This paper proposes a Visual Analytics process model for prototype learning, and, based on this, presents ProtoExplorer, a Visual Analytics system for the exploration and refinement of prototype-based deepfake detection models. ProtoExplorer offers tools for visualizing and temporally filtering prototype-based predictions when working with video data. It disentangles the complexity of working with spatio-temporal prototypes, facilitating their visualization. It further enables the refinement of models by interactively deleting and replacing prototypes with the aim to achieve more interpretable and less biased predictions while preserving detection accuracy. The system was designed with forensic experts and evaluated in a number of rounds based on both open-ended think aloud evaluation and interviews. These sessions have confirmed the strength of our prototype based exploration of deepfake videos while they provided the feedback needed to continuously improve the system.

Key findings

Evaluation with forensic experts showed ProtoExplorer effectively supports deepfake analysis. The system enables interactive model refinement, leading to more interpretable predictions without significant accuracy loss. Experts highlighted the need for improved prototype candidate selection and visualizations, suggesting future research directions.

Approach

ProtoExplorer uses a visual analytics approach to improve the interpretability of prototype-based deepfake detection. It allows users to visualize spatio-temporal prototypes, filter predictions temporally, and interactively delete or replace prototypes to optimize both interpretability and accuracy.

Datasets

FaceForensics++ (HQ variant, using Deepfakes, Neural Textures, FaceSwap, and Face2Face deepfake generation techniques)

Model(s)

Dynamic Prototype Network (DPNet) with HRNet as a feature encoder

Author countries

Netherlands

← Previous