Abstract: Systems and methods for code analysis are provided. A system receives, from a first data owner system, an input indicating one or more privacy constraints, wherein the one or more privacy constraints indicate one or more rules for operating on a sensitive data set in a data store associated with the first data owner system. The system receives data comprising code configured to operate on the sensitive data set. The system apples a first set of one or more machine-learning-trained models to process the first input and the data comprising the code to generate code analysis output data, wherein the code analysis output data comprises an indication of whether the code satisfies the one or more privacy constraints. The system generates and transmits, based on the code analysis output data, an instruction indicating whether to execute the code to operate on the sensitive data.

Abstract: Provided herein are systems and methods for vertical federated machine learning. Vertical federated machine learning can be performed by a central system communicatively coupled to a plurality of satellite systems. The central system can receive encrypted data from the satellite systems and apply a transformation that transforms the encrypted data into transformed data. The central system can identify matching values in the transformed data and generate a set of location indices that indicate one or more matching values in the transformed data. The central system can transmit instructions to the satellite systems to access data stored at locations indicated by the location indices and to train a machine learning model using data associated with said locations.

Abstract: Provided herein are systems and methods for federated machine learning performed by a central system communicatively coupled to plurality of satellite systems that implement privacy-preserving techniques. Synthetic data generated at respective satellite systems based on the actual data of the satellite systems can be utilized to generate data processing rules that can be applied to the actual data and used to develop a central machine learning model. The systems and methods disclosed herein can be used for both horizontal or vertical federated machine learning by implementing an alignment algorithm as necessary. Insights based on synthetic data and/or the alignment algorithm can be used to develop a central machine learning model without accessing any actual data values directly. Local models can be generated by training the central machine learning model at respective satellite sites and then aggregated at the central system, without transmitting the actual data from the respective satellite systems.

Abstract: Systems and methods for private deep neural network training are disclosed. Method includes storing first private values at first machine and second private values at second machine; providing, to third machine, first share of first private values and first share of second private values; providing, to fourth machine, second share of first private values and second share of second private values; computing, at third machine, third machine-value based on first share of first private values and first share of second private values; computing, at fourth machine, fourth machine-value based on second share of first private values and second share of second private values; providing, to first machine and second machine, third machine-value and fourth machine-value; and computing, at first machine, a mathematical function of first private values and second private values, mathematical function being computed based on first private values stored at first machine, third machine-value, and fourth machine-value.

Abstract: Systems and methods for private deep neural network training are disclosed. Method includes storing first private values at first machine and second private values at second machine; providing, to third machine, first share of first private values and first share of second private values; providing, to fourth machine, second share of first private values and second share of second private values; computing, at third machine, third machine-value based on first share of first private values and first share of second private values; computing, at fourth machine, fourth machine-value based on second share of first private values and second share of second private values; providing, to first machine and second machine, third machine-value and fourth machine-value; and computing, at first machine, a mathematical function of first private values and second private values, mathematical function being computed based on first private values stored at first machine, third machine-value, and fourth machine-value.

Abstract: An approach to implementing or configuring an Oblivious RAM (ORAM), which in addition to behaving as a RAM, provides a way to meet a specified degree of privacy in a manner that avoids applying unnecessary computation resources (computation time and/or storages space and/or data transfer) to achieve the specified degree of privacy. In this way, a tradeoff between privacy and computation resources may be tuned to address requirements of a particular application. This ability to tune this tradeoff is not found in other ORAM implementations, which in general aim to achieve complete privacy. In some implementations, the ORAM provides a constant bandwidth overhead compared to conventional RAMs, while achieving a statistical privacy as desired by the user.

Abstract: An approach to implementing or configuring an Oblivious RAM (ORAM), which in addition to behaving as a RAM, provides a way to meet a specified degree of privacy in a manner that avoids applying unnecessary computation resources (computation time and/or storages space and/or data transfer) to achieve the specified degree of privacy. In this way, a tradeoff between privacy and computation resources may be tuned to address requirements of a particular application. This ability to tune this tradeoff is not found in other ORAM implementations, which in general aim to achieve complete privacy. In some implementations, the ORAM provides a constant bandwidth overhead compared to conventional RAMs, while achieving a statistical privacy as desired by the user.