# Sparse deepfake detection promotes better disentanglement

arXiv: https://arxiv.org/abs/2510.05696 · HTML: https://mlsanity.com/paper/2510.05696/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 17 of its 25 references that are in the map.
- First posted on arXiv 7 Oct 2025.

## Builds on (most cited first, 17 of 17)

- [Attention is not not Explanation](https://mlsanity.com/paper/1908.04626/) (2019)
- [Scaling and evaluating sparse autoencoders](https://mlsanity.com/paper/2406.04093/) (2024)
- [MLS: A Large-Scale Multilingual Dataset for Speech Research](https://mlsanity.com/paper/2012.03411/) (2020)
- [AASIST: Audio Anti-Spoofing using Integrated Spectro-Temporal Graph Attention Networks](https://mlsanity.com/paper/2110.01200/) (2021)
- [Comparative layer-wise analysis of self-supervised speech models](https://mlsanity.com/paper/2211.03929/) (2022)
- [ASVspoof 5: Crowdsourced Speech Data, Deepfakes, and Adversarial Attacks at Scale](https://mlsanity.com/paper/2408.08739/) (2024)
- [Measuring Disentanglement: A Review of Metrics](https://mlsanity.com/paper/2012.09276/) (2020)
- [SPINE: SParse Interpretable Neural Embeddings](https://mlsanity.com/paper/1711.08792/) (2017)
- [Probing phoneme, language and speaker information in unsupervised speech representations](https://mlsanity.com/paper/2203.16193/) (2022)
- [Explaining deep learning models for spoofing and deepfake detection with SHapley Additive exPlanations](https://mlsanity.com/paper/2110.03309/) (2021)
- [Malafide: a novel adversarial convolutive noise attack against deepfake and spoofing detection systems](https://mlsanity.com/paper/2306.07655/) (2023)
- [DINE: Dimensional Interpretability of Node Embeddings](https://mlsanity.com/paper/2310.01162/) (2023)
- [Malacopula: adversarial automatic speaker verification attacks using a neural-based generalised Hammerstein model](https://mlsanity.com/paper/2408.09300/) (2024)
- [Towards Explainable Spoofed Speech Attribution and Detection:a Probabilistic Approach for Characterizing Speech Synthesizer Components](https://mlsanity.com/paper/2502.04049/) (2025)
- [BanditPAM: Almost Linear Time k-Medoids Clustering via Multi-Armed Bandits](https://mlsanity.com/paper/2006.06856/) (2020)
- [Unsupervised Multiple Domain Translation through Controlled Disentanglement in Variational Autoencoder](https://mlsanity.com/paper/2401.09180/) (2024)
- [Explaining Speaker and Spoof Embeddings via Probing](https://mlsanity.com/paper/2412.18191/) (2024)

## Abstract (arXiv)

Due to the rapid progress of speech synthesis, deepfake detection has become a major concern in the speech processing community. Because it is a critical task, systems must not only be efficient and robust, but also provide interpretable explanations. Among the different approaches for explainability, we focus on the interpretation of latent representations. In such paper, we focus on the last layer of embeddings of AASIST, a deepfake detection architecture. We use a TopK activation inspired by SAEs on this layer to obtain sparse representations which are used in the decision process. We demonstrate that sparse deepfake detection can improve detection performance, with an EER of 23.36% on ASVSpoof5 test set, with 95% of sparsity. We then show that these representations provide better disentanglement, using completeness and modularity metrics based on mutual information. Notably, some attacks are directly encoded in the latent space.