Some embodiments are directed to location-tracking system (100) comprising a location database (120) configured to receive a plurality of location updates from a plurality of tracking devices (112, 113), the plurality of location updates indicating the location of one or more objects, the location updates being stored encrypted with a cryptographic database encryption-key (130), multiple location-analysis devices execute a multi-party computation protocol on the encrypted location updates using a stored key-share, thus jointly computing a location-analysis result secret-shared among the multiple location analysis devices.

An encryption system and method that addresses private computation in public clouds and provides the ability to perform operations of encrypted data (including equality determinations and compare for less than operations) are provided.

Disclosed is a method that includes training, at a client, a part of a deep learning network up to a split layer of the client. Based on an output of the split layer, the method includes completing, at a server, training of the deep learning network by forward propagating the output received at a split layer of the server to a last layer of the server. The server calculates a weighted loss function for the client at the last layer and stores the calculated loss function. After each respective client of a plurality of clients has a respective loss function stored, the server averages the plurality of respective weighted client loss functions and back propagates gradients based on the average loss value from the last layer of the server to the split layer of the server and transmits just the server split layer gradients to the respective clients.

Disclosed is a process for testing a suspect model to determine whether it was derived from a source model. An example method includes receiving, from a model owner node, a source model and a fingerprint associated with the source model, receiving a suspect model at a service node, based on a request to test the suspect model, applying the fingerprint to the suspect model to generate an output and, when the output has an accuracy that is equal to or greater than a threshold, determining that the suspect model is derived from the source model. Imperceptible noise can be used to generate the fingerprint which can cause predictable outputs from the source model and a potential derivative thereof.

According to one aspect, there is provided a server for use in evaluating a monitoring function to determine if a trigger condition is satisfied. The server comprises a processing unit and a memory unit. The memory unit is for storing a current monitoring state (Ss) of the server or an encrypted current monitoring state (S) of the monitoring function, the current monitoring state (Ss) of the server relating to the current monitoring state (S) of the monitoring function that is based on an evaluation of one or more previous events. The processing unit is configured to receive an indication of a first event from a first client node and evaluate the monitoring function to determine if the first event satisfies the trigger condition. The evaluation is performed using a privacy-preserving computation (PPC), with the server providing the current monitoring state (Ss) of the server as a first private input to the PPC or the encrypted current monitoring state (S) of the monitoring function as a first input to the PPC, and the first client node providing the first event or an encryption thereof as a private input to the PPC. The evaluation of the monitoring function provides an encrypted updated monitoring state (S′) of the monitoring function or an updated monitoring state (Ss′) of the server as an output of the monitoring function and an indication of whether the first event satisfies the trigger condition.

According to an aspect, there is provided a first node for evaluating an event using a function. A corresponding computer-implemented method of operating a first node to 5 evaluate an event using a function is also provided. The function is evaluated by two parties using garbled circuits, with each party garbling a circuit representing the function, and evaluating the circuit garbled by the other party.

The present approaches generally relate to the encryption of data within a database in such a way that the encrypted data may still be easily accessed and utilized by an application. The present approach provides the ability to encrypt and decrypt data at an application layer though the data remains in an encrypted state at the database layer and when in transit.

Malicious website detection has been very crucial in timely manner to avoid phishing. User privacy also needs to be maintained at the same time. A system and method for classifying a website URL have been provided. The system is configured to achieve end-to-end privacy for machine learning based malicious URL detection. The system provides privacy preserving malicious URL detection models based on Fully Homomorphic Encryption (FHE) approach either using deep neural network (DNN), using logistic regression or using a hybrid approach of both. The system is utilizing a split architecture (client-server) where-in feature extraction is done by a client machine and classification is done by a server. The client machine encrypts the query using FHE and sends it to the server which hosts machine learning model. During this process, the server doesn't learn any information about the query.

There is provided a requestor device for digital signing of a message, comprising: at least one hardware processor executing a code for: transmitting the message for signing thereof, in a single request session over the network to each one of a plurality of validator devices, wherein a beacon device computes and transmits over a network to each one of a plurality of validator devices a signature-data value computed and signed by the beacon device, receiving in a single response session from each one of the plurality of validator devices, a respective partial-open decrypted value computed for the signature-data value and the message, and aggregating the partial-opens decrypted values received from the plurality of validator devices to compute the digital signature of the message.

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.