語音複製與深偽音訊

# 範例：以 XTTS（Coqui TTS）進行語音複製
from TTS.api import TTS
 
def clone_voice(
    reference_audio_path: str,
    text_to_speak: str,
    output_path: str = "cloned_output.wav"
) -> str:
    """自參考音訊複製語音並產生新語音。"""
    tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2")
 
    tts.tts_to_file(
        text=text_to_speak,
        speaker_wav=reference_audio_path,
        language="en",
        file_path=output_path
    )
    return output_path

# 重播本身很簡單——挑戰在於繞過活體偵測
# 有些系統會檢查：
# 1. 背景雜訊模式（太乾淨即可疑）
# 2. 麥克風特徵
# 3. 即時互動（隨機挑戰片語）

def bypass_text_dependent_verification(
    target_voice_sample: str,
    challenge_phrase: str
) -> str:
    """
    以目標聲音生成挑戰片語。
    可繞過要求使用者朗讀特定片語之文字相依驗證。
    """
    return clone_voice(
        reference_audio_path=target_voice_sample,
        text_to_speak=challenge_phrase,
        output_path="bypass_attempt.wav"
    )

def adversarial_speaker_embedding(
    speaker_model,
    target_embedding: torch.Tensor,
    source_audio: torch.Tensor,
    num_steps: int = 500
) -> torch.Tensor:
    """
    修改來源音訊以吻合目標語者 embedding，
    同時保留所說內容。
    """
    delta = torch.zeros_like(source_audio, requires_grad=True)
 
    for step in range(num_steps):
        adv_audio = source_audio + delta
        current_embedding = speaker_model.encode(adv_audio)
 
        # 最小化與目標 embedding 之距離
        loss = torch.nn.functional.mse_loss(
            current_embedding, target_embedding
        )
        loss.backward()
 
        with torch.no_grad():
            delta.data -= 0.001 * delta.grad.sign()
            delta.data = torch.clamp(delta.data, -0.05, 0.05)
 
        delta.grad.zero_()
 
    return (source_audio + delta).detach()

import torch
import numpy as np
 
def extract_deepfake_features(audio: np.ndarray, sr: int = 16000) -> dict:
    """
    擷取合成音訊的指標特徵。
 
    真實語音具有難以完美複製的特徵：
    - 基頻微變（jitter）
    - 振幅波動（shimmer）
    - 自然呼吸模式
    - 共振峰轉換
    """
    features = {}
 
    # 音高 jitter（基頻的變動）
    # 合成聲音常具不自然平滑的音高
    # 此為簡化檢查
    frame_size = int(0.03 * sr)  # 30ms frame
    energies = []
    for i in range(0, len(audio) - frame_size, frame_size):
        frame = audio[i:i + frame_size]
        energies.append(np.sqrt(np.mean(frame ** 2)))
 
    features["energy_variance"] = np.var(energies)
    features["energy_jitter"] = np.mean(np.abs(np.diff(energies)))
 
    # 頻譜平坦度（合成音訊常具不同之頻譜特性）
    from scipy.fft import rfft
    spectrum = np.abs(rfft(audio))
    geometric_mean = np.exp(np.mean(np.log(spectrum + 1e-10)))
    arithmetic_mean = np.mean(spectrum)
    features["spectral_flatness"] = geometric_mean / (arithmetic_mean + 1e-10)
 
    return features