Machine Learning Mitigants for Speech Based Cyber Risk

Statistical analysis of speech is an emerging area machine learning. In this paper, we tackle the biometric challenge Automatic Speaker Verification (ASV) differentiating between samples generated by two distinct populations utterances, those authentic human voice and a synthetic one. Solving such issue through statistical perspective foresees definition decision rule function learning procedure to identify optimal classifier. Classical state-of-the-art countermeasures rely on strong assumptions as stationarity or local-stationarity that may be atypical encounter in practice. We explore regard robust non-linear non-stationary signal decomposition method known Empirical Mode Decomposition combined with Mel-Frequency Cepstral Coefficients novel fashion refined classifier technique multi-kernel Support Vector machine. undertake significant real data case studies covering multiple ASV systems using different datasets, including ASVSpoof 2019 database. The obtained results overwhelmingly demonstrate significance our feature extraction approach versus existing conventional methods reducing threat cyber-attack perpetrated replication seeking unauthorised access.