selfcontrol7/Korean_Voice_Phishing_Detection

Implements multiple detection paradigms—traditional ML classifiers, CNN-BiLSTM hybrids with attention mechanisms, KoBERT language models, and federated learning—all trained on the KorCCVi dataset of Korean voice phishing transcripts. The repository provides modular implementations across preprocessing, model training, and evaluation pipelines, enabling researchers to compare architectural approaches from classical to transformer-based methods. Targets reproducibility through documented paper implementations with citation support and Binder integration for interactive experimentation.

No commits in the last 6 months.