A
privacy preserving computation protocol for data analytics is described. The protocol includes a method for privacy-preserving computation of aggregated private data of a group of
client devices wherein the method comprises: a
server selecting at least t
client devices from the group of
client devices, each client device in the group: being identifiable by client index i; comprising an
encryption function; being provided with key information including an
encryption key e and a decryption key of a homomorphic threshold
cryptosystem; generating or being provided with an random value ri and having access to or being provided with the random values of the other client devices in the group; the
server transmitting client information to each selected client device, the client information including client indices identifying the selected client devices, the client information signalling a client device that the
server would like aggregate encrypted private data of each of the selected client devices; the server receiving randomized encrypted private data and an associated decryption share from each selected client device, the decryption shares being configured such that decryption key d can be reconstructed on the basis of t decryption shares; and, the server aggregating, preferably summing or adding, the received randomized encrypted private data of the selected client devices using the homomorphic properties of the
cryptosystem and using the decryption shares for decrypting the aggregated randomized encrypted private data into cleartext.