CAJ | 학술논문

回归关系推断是统计分析领域中的重要基础内容,其运算形式较为复杂,导致了安全解决方案的效率低下.为了解决安全双方模型下多元回归的统计推断问题,提出极大似然估计、残差平方和与总平方和的安全协议.并依据运算本身的特点,经过对共享矩阵连乘操作规律的数学分析,提出简化运算的方法,使得交互次数大幅减少.同时对其可应用领域进行了展望.
회귀관계추단시통계분석영역중적중요기출내용,기운산형식교위복잡,도치료안전해결방안적효솔저하.위료해결안전쌍방모형하다원회귀적통계추단문제,제출겁대사연고계、잔차평방화여총평방화적안전협의.병의거운산본신적특점,경과대공향구진련승조작규률적수학분석,제출간화운산적방법,사득교호차수대폭감소.동시대기가응용영역진행료전망.
Inference of regression relationship is a fundamental content in statistical analysis. Its operational form is somehow too complicated to be computed efficiently. In order to solve multiple linear regression under secure two-party model, the secure protocols for maximum likelihood, sum of squared residuals and regression sum of square are proposed. According to their specialties of calculation , the simplification discipline of matrix continued product is summarized and the number of interactive time is reduced. At last, the applications for privacy-preserving predication detection are also discussed.