There is provided a secure computing server that performs shift operation on secretly distributed shares. The secure computing server may perform the shift operation when a number of significant digits of secret information corresponding to a secretly distributed share is to be reduced.

Collaborative multiparty homomorphic encryption comprising receiving a linear common public key collaboratively generated by a plurality of parties as a sum of linear public key shares associated with the respective plurality of parties. Each of two ciphertexts may be encrypted with the linear common public key and the two ciphertexts may be combined by a non-linear computation to generate a result ciphertext encrypted by a non-linear public key. The result ciphertext may be re-encrypted with a re-linearization key to swap encryption keys from the non-linear public key to a linear public key. The re-encrypted result ciphertext may be distributed to the plurality of parties to each partially decrypt the re-encrypted result ciphertext by a linear secret key share associated with the party, which in combination fully decrypts the result by a linear common secret key that is a sum of the secret key shares of the respective plurality of parties.

A device, system and method for privacy enhanced proximity detection by secure collaboration between a first party without access to user locations and a second party without access to a target user identifier. The second party may receive from the first party a homomorphic encryption public key and homomorphic encrypted target user identifier or masked target location, and may determine an associated homomorphic encrypted target user location. The second party may search a homomorphically encrypt database of user locations and associated user identifiers for homomorphic encrypted proximate user identifiers associated with homomorphic encrypted user locations proximate to the homomorphic encrypted target user location. The second party may send the first user the search result of homomorphic encrypted proximate user identifiers to be decrypted by the first party with a private key to identify proximate user identifiers without knowing their locations.

Traditional deep learning techniques are performed by high-performance system with direct access to the data to train large models. An approach of training the model from a collaboration of similar stakeholders where they pool together their data in a central server. However, data privacy is lost by exposing said models and data security while accessing heterogeneous data. Embodiments of the present disclosure provide a method and system for a cross-silo serverless collaborative learning among a plurality of clients in a malicious client threat-model based on a decentralized Epsilon cluster selection. Protocols are initialized and considered to iteratively train local models associated with each client and aggregate the local models as private input based on the multi-party computation to obtain global model. Non-linear transformation of a silhouette score to an Epsilon probability without implementing a server to select r.sup.th model from an active set to assign as the global model.

Data storage nodes that participate in a requested data statistical analysis as participant data storage nodes are determined and divided into a plurality of node sets. Data stored in each participant data storage node associated with a particular node set is encrypted, where the encrypted data is divided into a number of fragments at least equal to a number of participant data storage nodes associated with the particular node set. Each participant data storage node sends a portion of the encrypted data to each of the other participant data storage nodes within the particular node set. Each participant data storage node processes received encrypted data and data remaining on the particular participant data storage node to obtain a processing result. Each participant data storage node sends the processing result to a proxy node, wherein the proxy node performs data statistical analysis based on the processing result.

Embodiments of a secure multi-party computation method applicable to any one computing node of a plurality of computing nodes deployed in a distributed network are provided. The plurality of computing nodes jointly participate in a secure multi-party computation based on private data held by each computing node. The computing node is connected to a trusted key source, and the method includes: obtaining a trusted key from the trusted key source; encrypting the private data held by the computing node based on the obtained trusted key to obtain ciphertext data; transmitting a computing parameter comprising at least the ciphertext data to other computing nodes participating in the secure multi-party computation, so that the other computing nodes perform the secure multi-party computation based on collected computing parameters transmitted by the computing nodes participating in the secure multi-party computation.

Several round-efficient solitary multi-party computation protocols with guaranteed output delivery are disclosed. A plurality of input devices and an output device can collectively perform a computation using methods such as fully homomorphic encryption. The output of the computation is only known to the output device. Some number of these devices may be corrupt. However, even in the presence of corrupt devices, the output device can still either generate a correct output or identify that the computation was compromised. These protocols operate under different assumptions regarding the communication infrastructure (e.g., broadcast vs point-to-point), the number of participating devices, and the number of corrupt devices. These protocols are round-efficient in that they require a minimal number of communication rounds to calculate the result of the multi-party computation.

Multiple elements are efficiently read from a secured array. A secure text array <a>=(<a[0]>, . . . , <a[n−1]>) where an array a=(a[0], . . . , a[n−1]) having a size of n is secured, secure text <x> of an integer x that is equal to or higher than 0 and less than n, and in integers i.sub.0, . . . , i.sub.m-1 that are equal to or higher than 0 and less than n are input into an input part 11. A secure shift part 12 secure-shifts the secure text array <a> by <x> to obtain a secure text array <a′>=(<a′[0]>, . . . , <a′[n−1]>) where an array a′=(a′[0], . . . , a′[n−1]) obtained by shifting leftward the array a by x is secured. An array generation part 13 generates a secure text array <b>=(<a′[i.sub.0]>, . . . , <a′[i.sub.m-1]>) from the secure text array <a′>.

A method for implementing a secure multiparty computation protocol between a plurality of parties for a multiparty computation includes performing an offline phase of an SPDZ protocol for each of the parties participating in the multiparty computation. A secret share redistribution phase is then performed wherein the secret shares of the parties are redistributed to a subset of the parties. A secret share recombination phase is performed during which the subset of the parties recombines the redistributed secret shares to recover the secret shares of the parties not in the subset. An online phase of the SPDZ protocol is then performed during which the function is computed with respect to the private inputs of the parties and using the secret shares of all the parties.

Disclosed herein are systems and methods for generating tokens using SMPC compute engines. In one aspect, a method may hash, by a node, a data input with a salt value. The method may split, by the node, the hashed data input into a plurality of secret shares, wherein each respective secret share of the plurality of secret shares is assigned to a respective SMPC compute engine of a plurality of SMPC compute engines. The respective SMPC compute engines may be configured to collectively hash the respective secret share with a secret salt value, unknown to the plurality of SMPC compute engines. The respective SMPC compute engine may further receive a plurality of hashed secret shares from remaining SMPC compute engines of the plurality of SMPC compute engines, and generate a token, wherein the token is a combination of the hashed respective secret share and the plurality of hashed secret shares.