Abstract: A computer-implemented method and a distributed computer system (100) for privacy-preserving distributed training of a global neural network model on distributed datasets (DS1 to DSn). The system has a plurality of data providers (DP1 to DPn) being communicatively coupled. Each data provider has a respective local training dataset (DS1 to DSn) and a vector of output labels (OL1 to OLn) for training the global model. Further, it has a portion of a cryptographic distributed secret key (SK1 to SKn) and a corresponding collective cryptographic public key (CPK) of a multiparty fully homomorphic encryption scheme, with the weights of the global model being encrypted with the collective public key. Each data provider (DP1) computes and aggregates, for each layer of the global model, encrypted local gradients (LG1) using the respective local training dataset (DS1) and output labels (OL1), with forward pass and backpropagation using stochastic gradient descent.

Abstract: A computer-implemented method and a distributed computer system (100) for privacy-preserving distributed training of a global model on distributed datasets (DS1 to DSn). The system has a plurality of data providers (DP1 to DPn) being communicatively coupled. Each data provider has a respective local model (LM1 to LMn) and a respective local training dataset (DS1 to DSn) for training the local model using an iterative training algorithm (IA). Further it has a portion of a cryptographic distributed secret key (SK1 to SKn) and a corresponding collective cryptographic public key (CPK) of a multiparty fully homomorphic encryption scheme, with the local and global model being encrypted with the collective public key. Each data provider (DP1) trains its local model (LM1) using the respective local training dataset (DS1) by executing gradient descent updates of its local model (LM1), and combining (1340) the updated local model (LM1?) with the current global model (GM) into a current local model (LM1c).

Abstract: Disclosed embodiments include methods, apparatuses, and systems for secured image processing, image recognition, biometric recognition, and face recognition in untrusted environments. The disclosure includes a system for secure image recognition that comprises a secure biometric recognition system configured to work directly with encrypted signals, and the secure biometric recognition system comprises an input quantization system and a homomorphic encryption system configured for noninteractive biometric recognition.

Abstract: Disclosed embodiments include a cryptographic system implemented in at least one digital computer with one or more processors or hardware such as FPGAs for performing iterative secure computations, analysis, and signal processing directly on encrypted data in untrusted environments. According to a basic embodiment, the proposed cryptographic system comprises: (a) at least one secure protocol for performing matrix multiplications in the encrypted domain, and (b) at least one secure iterative protocol for solving systems of linear equations in the encrypted domain. According to a particular embodiment the system comprises a plurality of privacy-preserving protocols for solving systems of linear equations (SLE) directly based on homomorphic computation and secret sharing. More specifically, according to a particular embodiment the system uses a protocol whereby systems of linear equations are solved securely and iteratively without imposing any restrictions on the matrix coefficients.

Abstract: Disclosed embodiments include methods and apparatuses for secure iterative processing of encrypted signals based on implementing a secure iterative processing protocol that avoids cipher blowup, and applying an iterative algorithm directly on the encrypted signals to generate an encrypted processed output signal. In a particular embodiment, the protocol comprises applying homomorphic linear processing, preparing and applying a rescaling module to avoid cypher blowup, and homomorphically adapting the encrypted signals. Specific embodiments implement iterative adaptive filtering algorithms in the encrypted domain including non-collaborative secure filtering and two-party filtering based on homomorphic encryption, interactive secure protocols, garbled circuits, or a combination thereof.

Abstract: Disclosed embodiments include a cryptographic system implemented in at least one digital computer with one or more processors or hardware such as FPGAs for performing secure computations, analysis, and signal processing directly on encrypted data in untrusted environments. According to a basic embodiment, the proposed cryptographic system comprises: (a) at least one secure protocol for performing matrix multiplications in the encrypted domain, and (b) at least one secure protocol for solving systems of linear equations in the encrypted domain. According to a particular embodiment, the system comprises a plurality of privacy-preserving protocols for solving systems of linear equations (SLE) directly based on homomorphic computation and secret sharing. More specifically, according to a particular embodiment, the system uses a protocol whereby systems of linear equations are solved securely by direct Gaussian elimination using a secure protocol without imposing any restrictions on the matrix coefficients.

Abstract: Disclosed embodiments include methods and apparatuses for secure iterative processing of encrypted signals based on implementing a secure iterative processing protocol that avoids cipher blowup, and applying an iterative algorithm directly on the encrypted signals to generate an encrypted processed output signal. In a particular embodiment, the protocol comprises applying homomorphic linear processing, preparing and applying a resealing module to avoid cypher blowup, and homomorphically adapting the encrypted signals. Specific embodiments implement iterative adaptive filtering algorithms in the encrypted domain including non-collaborative secure filtering and two-party filtering based on homomorphic encryption, interactive secure protocols, garbed circuits, or a combination thereof.

Abstract: Disclosed embodiments include a cryptographic system implemented in at least one digital computer with one or more processors or hardware such as FPGAs for performing iterative secure computations, analysis, and signal processing directly on encrypted data in untrusted environments. According to a basic embodiment, the proposed cryptographic system comprises: (a) at least one secure protocol for performing matrix multiplications in the encrypted domain, and (b) at least one secure iterative protocol for performing matrix inversions and solving systems of equations based on an iterative secure protocol substantially equivalent to a Newton secure protocol. According to a particular embodiment, the system comprises a plurality of privacy-preserving protocols for solving systems of linear equations (SLE) directly based on homomorphic computation and secret sharing.

Abstract: Disclosed embodiments include a cryptographic system implemented in at least one digital computer with one or more processors or hardware such as FPGAs for performing iterative secure computations, analysis, and signal processing directly on encrypted data in untrusted environments. According to a basic embodiment, the proposed cryptographic system comprises: (a) at least one secure protocol for performing matrix multiplications in the encrypted domain, and (b) at least one secure iterative protocol for solving systems of linear equations in the encrypted domain. According to a particular embodiment the system comprises a plurality of privacy-preserving protocols for solving systems of linear equations (SLE) directly based on homomorphic computation and secret sharing. More specifically, according to a particular embodiment the system uses a protocol whereby systems of linear equations are solved securely and iteratively without imposing any restrictions on the matrix coefficients.

Abstract: Disclosed embodiments include a cryptographic system implemented in at least one digital computer with one or more processors or hardware such as FPGAs for performing secure computations, analysis, and signal processing directly on encrypted data in untrusted environments. According to a basic embodiment, the proposed cryptographic system comprises: (a) at least one secure protocol for performing matrix multiplications in the encrypted domain, and (b) at least one secure protocol for solving systems of linear equations in the encrypted domain. According to a particular embodiment, the system comprises a plurality of privacy-preserving protocols for solving systems of linear equations (SLE) directly based on homomorphic computation and secret sharing. More specifically, according to a particular embodiment, the system uses a protocol whereby systems of linear equations are solved securely by direct Gaussian elimination using a secure protocol without imposing any restrictions on the matrix coefficients.