Search papers, labs, and topics across Lattice.
This paper introduces an evolutionary multi-objective optimization framework using NSGA-II to fuse deepfake speech detectors, simultaneously minimizing detection error and system complexity. They explore binary-coded detector selection for score averaging and real-valued weights for weighted sum fusion. Experiments on ASVspoof 2015 using 36 SSL-based detectors demonstrate Pareto-optimal solutions that outperform simple averaging and logistic regression, achieving state-of-the-art performance with significantly reduced model size.
You can halve the size of your deepfake detection ensemble without sacrificing accuracy by using evolutionary multi-objective optimization to select and weight individual detectors.
While deepfake speech detectors built on large self-supervised learning (SSL) models achieve high accuracy, employing standard ensemble fusion to further enhance robustness often results in oversized systems with diminishing returns. To address this, we propose an evolutionary multi-objective score fusion framework that jointly minimizes detection error and system complexity. We explore two encodings optimized by NSGA-II: binary-coded detector selection for score averaging and a real-valued scheme that optimizes detector weights for a weighted sum. Experiments on the ASVspoof 5 dataset with 36 SSL-based detectors show that the obtained Pareto fronts outperform simple averaging and logistic regression baselines. The real-valued variant achieves 2.37% EER (0.0684 minDCF) and identifies configurations that match state-of-the-art performance while significantly reducing system complexity, requiring only half the parameters. Our method also provides a diverse set of trade-off solutions, enabling deployment choices that balance accuracy and computational cost.
