ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Conference Paper Comparing Self-Supervised and Domain-Invariant Features for Cross-Domain Voice Phishing Detection
Cited - time in scopus Share share facebook twitter linkedin kakaostory
Authors
Jeongmin Lee, Seung Yun, Minkyu Lee, Ran Han, Yoonkyu Woo, Jinxia Huang
Issue Date
2026-09
Citation
International Speech Communication Association (INTERSPEECH) 2026, pp.744-748
Publisher
ISCA
Language
English
Type
Conference Paper
DOI
https://dx.doi.org/10.21437/Interspeech.2026-1975
Abstract
Voice phishing detection faces three critical challenges: real criminal recordings are unavailable due to privacy constraints; when available, only a handful of samples exist, insufficient for fine-tuning; and lightweight acoustic-only detection is needed as an alternative to large self-supervised models. We compare domain-invariant prosodic features and self-supervised representations (HuBERT, wav2vec2.0) through cross-domain evaluation—training on scenario-based actor recordings and testing on authentic criminal calls. Domain-invariant prosodic features achieve 69.5% F1 zero-shot and 71.0% with 5-shot learning. HuBERT achieves highest performance (94.2% F1, 5-shot), while wav2vec2.0 exhibits a precision-oriented detection profile (90.2% F1 with 99.4% precision, 5-shot). These findings reveal fundamental trade-offs: domain-invariant features enable zero-shot deployment when no real data exists, while SSL methods achieve higher performance but require real samples and compute.
Keyword
voice phishing detection, cross-domain speech classification, self-supervised learning, prosodic features, few-shot learning
KSP Keywords
Cross-Domain, Higher performance, Phishing detection, Prosodic Features, Real data, Real samples, Scenario-based, Speech classification, Trade-off, Zero-shot, fine-tuning