GRIDEX: Grid-Grounded Forensic Explanations for Deepfake Spectrogram Analysis

2026-06-17

AI Summary

This work addresses the lack of interpretability in existing deepfake speech detection models, which hinders their ability to provide forensically admissible evidence with precise localization and clear acoustic meaning. To this end, the authors propose GRIDEX, a two-stage framework that first identifies critical anomalous regions in spectrograms and then generates structured forensic explanations encompassing time-domain, frequency-domain, phonetic, and semantic attributes. GRIDEX is the first method to enable region-localized, interpretable deepfake analysis by effectively linking localized forgery cues with contextual acoustic properties. By integrating supervised fine-tuning (SFT) with grouped relative policy optimization (GRPO) and leveraging a vision–language model for region selection and explanation generation, GRIDEX substantially outperforms strong VLM baselines on a newly curated dataset, achieving significant improvements in both forgery trace localization accuracy and explanation quality.

Abstract

The advancement of speech generation technologies has made artificial speech increasingly realistic. Although modern classification models can achieve high accuracy when it comes to deepfake detection, they do not produce evidences such as indicating where spoof cues appear in the spectrogram and what they imply acoustically, limiting their usefulness in forensic settings. Manual analysis of full spectrograms is resource-intensive, so evidence should narrow attention to the most diagnostic regions. Moreover, existing explainability methods have limited capabilities in connecting contextual attributes to localized evidence, making explanations harder to verify. To overcome this limitation, we propose GRIDEX, a pipeline that, when given a deepfake spectrogram, generates forensic explanations of its anomalies. The pipeline (i) selects top-K anomalous regions in the spectrogram and (ii) produces an explanation for each anomaly. The explanations follow a schema of categorical acoustic fields, including temporal, spectral, phonetic information and interpretation text. To our knowledge, this is the first framework to generate structured forensic explanations using regional grounding for deepfake spectrograms. GRIDEX is trained with a two-stage learning paradigm that combines supervised fine-tuning (SFT) with Group Relative Policy Optimization (GRPO). Experiments on our dataset show improved artifact localization and explanation quality over strong vision-language model (VLM) baselines. The dataset and code will be released upon publication.

Problem

• deepfake detection
• forensic explanation
• spectrogram analysis
• explainable AI
• anomaly localization

Innovation

• grid-grounded explanation
• deepfake spectrogram analysis
• forensic interpretability
• anomaly localization
• structured acoustic explanation