CAJ | 학술논문

结合声纹特征和VQ（Vector Quantization）识别算法的特点，在对整数同态加密算法进行改进的基础上，提出基于同态加密（homomorphic encryption scheme）的声纹模板保护方法。将实数形式的码本和认证声纹特征转化为整数，采用改进的整数同态加密算法对数据加密并计算密文的欧氏距离分量差，运算结果经解密后用于计算平均最小量化误差最终进行决策输出。该方法克服了实数同态加密算法密文泄漏明文的大小关系、小数和正负数信息等缺陷，并能抵抗已知明文攻击。实验结果表明，加密后系统的身份认证性能可基本保持，验证了方法的有效性。
결합성문특정화VQ（Vector Quantization）식별산법적특점，재대정수동태가밀산법진행개진적기출상，제출기우동태가밀（homomorphic encryption scheme）적성문모판보호방법。장실수형식적마본화인증성문특정전화위정수，채용개진적정수동태가밀산법대수거가밀병계산밀문적구씨거리분량차，운산결과경해밀후용우계산평균최소양화오차최종진행결책수출。해방법극복료실수동태가밀산법밀문설루명문적대소관계、소수화정부수신식등결함，병능저항이지명문공격。실험결과표명，가밀후계통적신빈인증성능가기본보지，험증료방법적유효성。
A voiceprint template protection method based on homomorphic encryption scheme is proposed, which takes the characteristics of speech features and Vector Quantization(VQ)speaker recognition algorithms into account. Voice-print features and codewords are transformed into non-negative integers. Components of Euclidean distance are calculated in the ciphertext domain by improved homomorphic encryption algorithm in the range of the integer. Decision is made according to the average minimum quantization error which is calculated with decrypting homomorphically component addition of VQ distance. The ciphers don’t leak the decimal fraction and sign information of plain text and can resist known plaintext attacks. The experimental results demonstrate the proposed method can keep the performance of authenti-cation system as the same as in the original feature domain.