Representation Loss Minimization with Randomized Selection Strategy for Efficient Environmental Fake Audio Detection

TL;DR

This paper proposes a randomized selection strategy to reduce the dimensionality of foundation model representations in environmental audio deepfake detection, maintaining or improving performance while significantly decreasing computational costs.

Contribution

It introduces a simple random selection method that outperforms SOTA dimensionality reduction techniques in preserving performance and reducing model complexity.

Findings

Random selection preserves or improves detection performance.

Reduces model parameters and inference time by over 50%.

Outperforms PCA, SVD, KPCA, and GRP in efficiency.

Abstract

The adaptation of foundation models has significantly advanced environmental audio deepfake detection (EADD), a rapidly growing area of research. These models are typically fine-tuned or utilized in their frozen states for downstream tasks. However, the dimensionality of their representations can substantially lead to a high parameter count of downstream models, leading to higher computational demands. So, a general way is to compress these representations by leveraging state-of-the-art (SOTA) unsupervised dimensionality reduction techniques (PCA, SVD, KPCA, GRP) for efficient EADD. However, with the application of such techniques, we observe a drop in performance. So in this paper, we show that representation vectors contain redundant information, and randomly selecting 40-50% of representation values and building downstream models on it preserves or sometimes even improves performance. We show that such random selection preserves more performance than the SOTA dimensionality reduction techniques while reducing model parameters and inference time by almost over half.