Comments on “Unconditionally Secure, Universally Composable Privacy Preserving Linear Algebra”

In this paper, we discuss the security of a distributed inner product (DIP) protocol [IEEE Trans. Inf. Forensics Security, 11.1, 59, (2016)]. We show information leakage in this protocol that does not happen in an ideal execution of the DIP functionality. In some scenarios, this information leakage enables one of the parties to completely learn the other party's input. We will give examples of such scenarios.