Abstract: This paper focuses on the privacy requirements of sensitive real-number data in practical multi-party applications, and proposes a scheme for fixed-point privacy-preserving computation using integer homomorphic encryption. The scheme maps real-number data in signed fixed-point representation to integers through domain translation, and then performs privacy-preserving computation on the translated integers using multi-party fully homomorphic encryption over integer. More importantly, to address the issue of decimal point drift in fixed-point privacy-preserving computation, this paper presents a random decimal digit generation algorithm and a decimal digit truncation algorithm, along with proofs and analyses of their correctness. For n participants, the communication complexity O(n^2) and computational complexity O(n^3) of this scheme are unaffected by the decimal place length γ. Therefore, compared to the scheme by Catrina et al., which has complexities of O(n^2γ) and O(n^3γ), the performance of this scheme does not degrade with increasing decimal place length. Experimental validation also demonstrates that this scheme has higher efficiency and better practicality.