Example systems and methods for biometric authentication that can bridge fuzzy extractors with deep learning and achieve the goals of preserving privacy and providing recoverability from zero are disclosed. Embeddings comprising a face or speaker embedding in a non-Hamming distance space can be processed to create a personal reliable bit map and a reliable locality-sensitive hash (LSH) for mapping the non-Hamming distance space to a Hamming distance space. A fuzzy extractor can be applied to create metadata that can be stored on a computing device. A secret can be recovered from the metadata and can be used for identification.

A physically unclonable function (PUF) device comprises a plurality of conductors, at least some of which are arranged so that they interact electrically and/or magnetically with one another. A media surrounds at least a portion of each of the conductors, and circuitry applys an electrical challenge signal to at least one of the conductors and for receiving an electrical output from at least one of the other conductors to generate an identifying response to the challenge signal that is unique to the device.

Aspects and features of a cryptosystem and authentication for the cryptosystem, and a method or process for the cryptosystem, are described. In one example, a method for cryptographic communications includes storing a secret key, generating a system randomization number, and encrypting a plain data package into an encrypted data package by application of the plain data package, the secret key, and the system randomization number to a system of equations for encryption. The system of equations can be a system of linearly dependent equations in one example. Among other benefits, the cryptosystem relies upon the system of linearly dependent equations and the system randomization number to provide additional strength against known-plaintext attacks, chosen-plaintext attacks, and other types of attacks. The system is more semantically secure and offers ciphertext indistinguishability in a new approach using the system of linearly dependent equations.

Systems, and associated methods, involving both a trusted and an untrusted device where sensitive data or keys are shared between those devices are disclosed. A disclosed method includes storing a key in a secure memory on a first device, receiving sensitive data via a user interface on a second device, generating a set of white box encryption instructions based on the key using a white box encryption generator on the first device, generating a complete data representation of the set of white box encryption instructions using a secure processor on the first device, transmitting the complete data representation from the first device to the second device, and encrypting the sensitive data using the complete data representation on the second device. The complete data representation is not Turing complete and is not executable with respect to the second device.

A method and a system of providing a quantum random number based on a quantum entropy source. The method includes generating a quantum random number and providing the quantum random number to a device. The generating of the quantum random number based on the quantum entropy source includes determining, by a management unit, whether quantum random numbers stored in a storage are insufficient; generating a quantum random number based on the quantum entropy source using a quantum random number generator provided in a production unit in response to a determination that the quantum random numbers are insufficient; verifying, by a verification unit, the generated quantum random number based on NIST SP800-90B and SP800-22; and in response to quality of the generated quantum number satisfying a criterion, storing the verified quantum random number in the storage.

Embodiments are disclosed for a method. The method includes validating training data that is provided for training a machine learning model using ordinary differential equations. The method further includes generating pre-processed training data from the validated training data by generating encrypted training data from the validated training data using homomorphic encryption and generating random noise based on the validated training data. The method also includes training the machine learning model adversarially with the pre-processed training data.

An arithmetic apparatus includes an interface and a circuity. The interface is connected to an information processing apparatus that is connected to a client apparatus and that processes data in an encrypted state. The circuitry acquires, from the information processing apparatus, encryption input data or encryption target data encrypted with a first encryption key. The circuitry decrypts the acquired, encryption input data or encryption target data with a first decryption key. Then, the circuitry executes a predetermined arithmetic operation on the decrypted arithmetic operation target data, encrypts data of an arithmetic operation result obtained by the predetermined arithmetic operation with the first encryption to key, and outputs the encrypted data of the arithmetic operation result to the information processing apparatus.

A system and a method are disclosed for securing sensitive data for transaction requests using tokenization and encryption. A secure transfer system secures sensitive information of transaction requests. The secure transfer system may receive a transaction request file and generate a modified transaction request file by tokenizing values in the received file. For each transaction request in the file, the system may store a representation of the untokenized values in a datastore in conjunction with an identifier of the transaction request. This identifier may be generated from the tokenized values. The secure transfer system may use the identifier to query the datastore for the representation of the untokenized values. The system may decrypt encrypted values in the representation to generate a transaction request file of detokenized values, which may be provided to an automated clearing house to fulfill the transaction requests.

A method for detecting relay attacks between two communication platforms, the method including: receiving, at a first communication platform, a first signal sent via a first communication channel from a second communication platform, the first signal including information about a challenge; receiving, at the first communication platform, a second signal sent via a second communication channel from the second communication platform, the second signal being a start clock; receiving, at the first communication platform, a third signal sent via the second communication channel from the second communication platform, the third signal including the challenge; outputting, from the first communication platform, a response to the challenge via the first communication channel to the second communication platform; and determining, at the second communication platform, whether a relay attack has occurred based on a time elapsed from when the start clock began to when the response is received at the second communication platform.

A communication system includes a server and a plurality of terminals. The server manages a public key and a private key. Each terminal includes computing circuitry that adds noise to first target data stored in the terminal, encrypts the first target data, to which the noise is added, using the public key, randomly chooses and determines, out of the server and different terminals, a first transmission destination of the first target data, transmits the encrypted first target data to the determined first transmission destination, receives second target data from a different terminal, randomly chooses and determines, out of the server and the different terminals, a second transmission destination of the received second target data, and transmits the received second target data to the second determined transmission destination. The server receives target data transmitted from the terminals and decrypt the received target data using the private key.