International Business Machines Corporation

Intranational Business Machines Corporation

International Business Machines Corproation

Allon Adir

Ehud Aharoni

Omri Soceanu

Lev Greenberg

Ramy Masalha

Tamer Salman

Roy Emek

Gil Eliezer Shurek

Oded Margalit

Fady Copty

Hayim Shaul

Nir Drucker

Eitan Marcus

Dmitry Pidan

Eyal Kushnir

Ariel Farkash

Amir Nahir

Boris Rozenberg

Gilad Ezov

Michael Mirkin

Oded Sofer

Itai Jaeger

Anatoly Koyfman

John Martin Ludden

Ronen Levy

Charles Meissner

Eyal Bin

Sigal Asaf

Rosa Miroshnikov

Roza Miroshnikov

Asaf Polakovski

Avi Ziv

Vitali Sokhin

Yoav Avraham Katz

Dov Murik

Michal Rimon

Omer Y Boehm

Laurent Fournier

Brad Lee Herold

Shady Copty

Michael Vinov

Pedro Martin-de-Nicolas

Guy Moshkowich

Elena Tsanko

Tzach Schechner

Kirill Shoikhet

Shimon Landa

Dimtry Krestyashyn

Maxim Golubev

Andrey Klinger

Amir Voskoboynik

A technique for privacy-preserving homomorphic inferencing using a neural network having an activation function, such as a non-linear high-degree polynomial. The network is trained to learn input features of an input feature vector together with their associated inverses. During inferencing, an encrypted data point is received at the network. The data point comprises an input feature vector that has been extended with a set of one or more additional feature values, the set of one or more additional feature values having been generated by applying a normalized inverse function to respective one or more features in the feature vector. Homomorphic inferencing is performed on the encrypted data point using the machine learning model to generate an encrypted result, which is then returned. By applying the normalized inverse function, the high-degree polynomial can use any value of an input feature during inferencing, whether the value is within or outside of a particular input range.

Non-linear approximation robust to input range of homomorphic encryption analytics

An embodiment appends, into a concatenated table, a second plurality of records in a second table to a first plurality of records in a first table. An embodiment sorts, according to each identification value in the concatenated table, the concatenated table. An embodiment generates, using an equality mask derived from each identification value in the sorted table, an intersection table, the intersection table comprising a record in the first plurality of records with a first identifier value matching a second identifier value in a record in the second plurality of records. An embodiment generates, using a not-in-intersection mask derived from the equality mask, a not-in-intersection table. An embodiment adds contents of the intersection table and contents of the not-in-intersection table together, resulting in a join table.

Database table joining under fully homomorphic encryption

Mechanisms are provided for performing a tournament-league comparison process of a computer function. A request is received to execute the computer function on a vector data structure, where a result of the computer function is provided by executing the tournament-league comparison process. The vector data structure comprises a plurality of values where each value corresponds to a vector slot. At least one iteration of a tournament comparison operation is executed to generate a first intermediate ciphertext and indicator matrix, where the first intermediate ciphertext comprises fewer vector slots than the at least one input vector data structure. A plurality of iterations of a league comparison operation are executed based on the first intermediate ciphertext and one or more second intermediate ciphertexts generated at each iteration of the league comparison operation. A final iteration of the league comparison operation is executed that outputs a final result of the tournament-league comparison process.

Tournament-league mechanisms for comparison based computer operations

An example system includes a processor to mask a ciphertext using four random elements to generate masked ciphertexts. The processor can send the masked ciphertexts to a server device. The processor can receive masked plaintexts from the server device. The processor can unmask the masked plaintexts using the four random elements to generate unmasked plaintexts.

Masking ciphertexts using random elements for secure decryption

A computer-implemented method for generating hash values to determine string similarity is disclosed. The computer-implemented method includes converting a first text string of a first data set into a first set of shingles. The computer-implemented method further includes determining a weight associated with each shingle in the first set of shingles based, at least in part, on a particular record field associated with a shingle. The computer-implemented method further includes generating, based on a hash function, a hash value for each shingle in the first set of shingles. The computer-implemented method further includes reducing the hash value generated for each shingle in the first set of shingles, based, at least in part on the weight associated with the shingle.

String similarity based weighted min-hashing

A method, apparatus and computer program product for privacy-preserving homomorphic inferencing. In response to receipt of encrypted data, a ciphertext of real numbers is generated. Each real number has an associated sign that is desired to be maintained. A mask is then identified, preferably via an iterative algorithm that works on a trial and error basis to locate an appropriate solution. The mask comprises set of values randomly distributed over a given positive range and that remain positive after encoding under a fixed-point arithmetic and with a low scale value. Under homomorphic encryption, the ciphertext is then multiplied by the mask to generate a result comprising values corresponding to the real numbers in the ciphertext and that maintain their associated signs. The result is provided as a response to the encrypted data.

Efficient random masking of values while maintaining their sign under fully homomorphic encryption (FHE)

Mechanisms are provided for fully homomorphic encryption enabled graph embedding. An encrypted graph data structure, having encrypted entities and predicates, is received and, for each encrypted entity, a corresponding set of entity ciphertexts is generated based on an initial embedding of entity features. For each encrypted predicate, a corresponding predicate ciphertext is generated based on an initial embedding of predicate features. A machine learning process is iteratively executed, on the sets of entity ciphertexts and the predicate ciphertexts, to update embeddings of the entity features of the encrypted entities and update embeddings of predicate features of the encrypted predicates, to generate a computer model for embedding entities and predicates. A final embedding is output based on the updated embeddings of the entity features and predicate features of the computer model.

Graph based embedding under fully homomorphic encryption

A method, a neural network, and a computer program product are provided that optimize training of neural networks using homomorphic encrypted elements and dropout algorithms for regularization. The method includes receiving, via an input to the neural network, a training dataset containing samples that are encrypted using homomorphic encryption. The method also includes determining a packing formation and selecting a dropout technique during training of the neural network based on the packing technique. The method further includes starting with a first packing formation from the training dataset, inputting the first packing formation in an iterative or recursive manner into the neural network using the selected dropout technique, with a next packing formation from the training dataset acting as an initial input that is applied to the neural network for a next iteration, until a stopping metric is produced by the neural network.

Neural network training with homomorphic encryption

A method comprising: receiving an input tensor having a shape defined by [n, . . . , n], where k is equal to a number of dimensions that characterize the input tensor; receiving tile tensor metadata comprising: a tile tensor shape defined by [t, . . . , t], and information indicative of an interleaving stride to be applied with respect to each dimension of the tile tensor; constructing an output tensor comprising a plurality of the tile tensors, by applying a packing algorithm which maps each element of the input tensor to at least one slot location of one of the plurality of tile tensors, based on the tile tensor shape and the interleaving stride, wherein the interleaving stride results in non-contiguous mapping of the elements of the input tensor, such that each of the tile tensors includes a subset of the elements of the input tensor which are spaced within the input tensor according to the interleaving stride.

Efficient convolution in an environment that enforces tiles

In response to identifying that a Single Instruction, Multiple Data (SIMD) operation has been instructed to be performed or has been performed by a Fully-Homomorphic Encryption (FHE) software on one or more original ciphertexts, performing the following steps: Performing the same operation on one or more original plaintexts, respectively, that are each a decrypted version of one of the one or more original ciphertexts. Decrypting a ciphertext resulting from the operation performed on the one or more original ciphertexts. Comparing the decrypted ciphertext with a plaintext resulting from the same operation performed on the one or more original plaintexts. Based on said comparison, performing at least one of: (a) determining an amount of noise caused by the operation, (b) determining whether unencrypted data underlying the one or more original ciphertexts has become corrupt by the operation, and (c) determining correctness of an algorithm which includes the operation.

Analysis and debugging of fully-homomorphic encryption

An example system includes a processor to receive, from a client device, a delete query requesting deletion of a row of in a fully homomorphically encrypted (FHE) database. The processor can store an identifier of the row to be deleted in a deletion queue, where the row is to be replaced with values of a row to be inserted from a received insertion query.

Deleting and inserting rows in fully homomorphically encrypted databases using queues

An example system includes a processor to receive, at a setup or sign-up, a first cipher including a biometric template transformed using a first transformation and encrypted using a secret key, a second cipher including a security vector encrypted using the secret key, a third cipher including the biometric template transformed using a second transformation and encrypted, and a fourth cipher including an encrypted second security vector. The processor can receive, at a runtime or sign-in, a fifth cipher and a sixth cipher. The processor can verify that the fifth cipher includes a second biometric template transformed using the first transformation and encrypted using the secret key and that the sixth cipher includes the second biometric template transformed using the second transformation by testing a format attribute of the transformation functions using comparisons of inner products. The processor can authenticate a user based on the Euclidean distance between the first cipher and the sixth cipher not exceeding a threshold in response to detecting that the transformations are verified.

Authentication using transformation verification

An efficient packing method that will optimize use of the homomorphic encryption ciphertext slots, trading-off size, latency, and throughput. Technology for working with tensors (multi-dimensional arrays) in a system that imposes tiles, that is, fixed-size vectors. An example of a system that imposes tiles are homomorphic encryption schemes, where each ciphertext encrypts a vector of some fixed size. The tensors are packed into tiles and then manipulated via operations on those tiles. Also, syntax for notation for describing packing details. This technology interprets the tiles as multi-dimensional arrays, and combines them to cover enough space to hold the tensor. An efficient summation algorithm can then sum over any dimension of this tile tensor construct that exists in the physical or logical addressing space of a computer data memory.

Packing arbitrary tensors into tiles with descriptive shapes

Embodiments may include techniques to prevent illegal ciphertexts using distance computations on homomorphic and/or functional encrypted templates while detecting whether the resulting distance does not meet requirements for validity. For example, a method may comprise receiving and storing enrollment information from a client computer system, the enrollment information comprising an encrypted enrollment template of authentication data and a plurality of randomly generated encrypted templates, receiving an additional encrypted template to be used to authenticate the user from the client computer system, authenticating the user by determining distances between the received additional encrypted template and the stored encrypted enrollment template and the each of the stored plurality of randomly generated encrypted templates, and determining that authentication is successful when the received additional encrypted template meets distance requirements relative to the stored encrypted enrollment template and the each of the stored plurality of randomly generated encrypted templates.

Preventing malformed ciphertext attacks on privacy preserving biometric authentication

Embodiments may provide distance computations on homomorphic and/or functional encrypted vectors while detecting whether the resulting distance has wrapped around due to the vectors having elements not in an allowed range. A method of user authentication processing may comprise receiving and storing enrollment information from a client computer system, the enrollment information comprising a template of authentication data and at least one additional encrypted vector, receiving an additional template to be used to authenticate the user from the client computer system, authenticating the user using the received additional template using the stored template and the stored at least one additional encrypted vector, and determining that authentication is successful when the received additional template matches the stored template and is valid based on the stored at least one additional encrypted vector.

Ensure valid range of values of a vector for distance calculations using homomorphic encryption or functional encryption

Growing community of inventors

Kiryat Tivon, Israel

Allon Adir