International Business Machines Corporation

The University of Bristol

Shai Halevi

Craig B Gentry

Rosario Gennaro

Nalini Kanta Ratha

Charanjit Singh Jutla

Karthik Nandakumar

Sharathchandra UmapathiRao Pankanti

Don Coppersmith

Nevenko Zunic

David R Safford

Stephen Michael Matyas, Jr

Michael Steiner

Mohammed Peyravian

Glen Alan Jaquette

Pankaj Rohatgi

Ronald Perez

Grant W Wagner

Paul Ashley Karger

Pau-Chen Cheng

Trent Ray Jaeger

Angela Marie Schuett

Nigel P Smart

Luke James O'Connor

Tal D Rabin

Zhen Liu

Anton Viktorovich Riabov

Genady Grabarnik

Paul Merrill Greco

Nagui Halim

Paul Merril Greco

Mark David Feblowitz

Kay Schwendimann Anderson

Vinod Vaikuntanathan

Guy M Cohen

Stephane Herman Maes

Lior Horesh

William Eric Hall

Hugo M Krawczyk

Sharathchandra Umapathirao Pankanti

Jeffrey Scott Sorensen

Ran Canetti

Nibel P Smart

ML model(s) are created and trained using training data from user(s) to create corresponding trained ML model(s). The training data is in FHE domains, each FHE domain corresponding to an individual one of the user(s). The trained machine learning model(s) are run to perform inferencing using other data from at least one of the user(s). The running of the ML model(s) determines results. The other data is in a corresponding FHE domain of the at least one user. Using at least the results, it is determined which of the following issues is true: the results comprise objectionable material, or at least one of the trained ML model(s) performs prohibited release of information. One or more actions are taken to take to address the issue determined to be true. Methods, apparatus, and computer program product are disclosed.

Searching over encrypted model and encrypted data using secure single-and multi-party learning based on encrypted data

A framework is provided in which a querying agency can request (via a query entity) encrypted data through a service provider from a data owning agency that stores encrypted data. The framework uses homomorphic encryption. The data may be gallery entities, and each of the elements in the framework operate on doubly-encrypted information. The service provider compares a representation of an encrypted query entity from the querying agency and representations of encrypted gallery entities from the data owning agency, resulting in doubly-encrypted values of a metric between corresponding compared representations. The querying agency gets result(s), based on the metric, which indicate whether it is probable the service provider has data similar to or the same as query data in the query entity. The elements have to perform communication in order for the querying agency or the data owning agency to get cleartext information corresponding to the query entity.

Secure matching and identification of patterns

A request is received for specific information that can be determined using data in a database on a first computer system. Either at least some of the data is encrypted or the request is encrypted. The first computer system does not have a decryption key to decrypt the encrypted data or request. The first computer system performs compressible HE operations on the data to determine compressed ciphertext(s) that correspond to the specific information. The operations include using a first uncompressed HE scheme and a second compressed HE scheme. The first HE scheme is used on the data to create other multiple ciphertexts and the second HE scheme is used on the other multiple ciphertexts to pack the other multiple ciphertexts into fewer ciphertexts that are compressed. Both the HE schemes use a same secret key. The first computer system sends a response including compressed ciphertext(s) corresponding to the specific information.

Compressible (F)HE with applications to PIR

An AI model is trained by determining insights for a sequence of computations used in the AI model. The sequence is applied to encrypted data and label pair(s), wherein computational details of each of the computations are defined. Information may also be committed for selected ones of the sequence of computations into a distributed database. The committed information may include computational details used in processing performed for the selected computations, and the distributed database may have a property that the committed information for each selected computation is linked with a verifiable signature of integrity with a previously committed computation in the sequence. Indication is received from an end-user computer system of selected computation(s). Computational details of the indicated selected computation(s) are sent toward the end-user computer system for use by the end-user computer system for verifying the indicated selected computation(s). The end-user computer system can verify the selected computation(s).

Encrypted data model verification

Respective sets of homomorphically encrypted training data are received from multiple users, each encrypted by a key of a respective user. The respective sets are provided to a combined machine learning model to determine corresponding locally learned outputs, each in an FHE domain of one of the users. Conversion is coordinated of the locally learned outputs in the FHE domains into an MFHE domain, where each converted locally learned output is encrypted by all of the users. The converted locally learned outputs are aggregated into a converted composite output in the MFHE domain. A conversion is coordinated of the converted composite output in the MFHE domain into the FHE domains of the corresponding users, where each converted decrypted composite output is encrypted by only a respective one of the users. The combined machine learning model is updated based on the converted composite outputs. The model may be used for inferencing.

Secure multi-party learning and inferring insights based on encrypted data

An approach is provided in which an information handling system loads a set of encrypted binary code into a processor that has been encrypted based upon a unique key of the processor. The processor includes an instruction decoder that transforms the set of encrypted binary code into a set of instruction control signals using the unique key. In turn, the processor executes a set of instructions based on the set of instruction control signals.

Malware resistant computer

A computer system receives a set of data encrypted by a homomorphic encryption transformation. The computer system performs machine learning operations using the encrypted set of data. The machine learning operations build, using homomorphic operations, a trained model of the data having a mapping between the encrypted data and output of the trained model. The model is stored for use for performing inferencing of other encrypted data to determine a corresponding output of the trained model. The computer system may perform inferencing of the other encrypted data at least by accessing the stored trained model and predicting by using the trained model a label in an encrypted format that corresponds to the other encrypted data. The computer system may send the label toward the client for the client to decrypt the label.

Learning and inferring insights from encrypted data

A request is received for specific information that can be determined using data in a database on a first computer system. Either at least some of the data is encrypted or the request is encrypted. The first computer system does not have a decryption key to decrypt the encrypted data or request. The first computer system performs compressible HE operations on the data to determine compressed ciphertext(s) that correspond to the specific information. The operations include using a first uncompressed HE scheme and a second compressed HEscheme. The first HE scheme is used on the data to create other multiple ciphertexts and the second HE scheme is used on the other multiple ciphertexts to pack the other multiple ciphertexts into fewer ciphertexts that are compressed. Both the HE schemes use a same secret key. The first computer system sends a response including compressed ciphertext(s) corresponding to the specific information.

In one exemplary embodiment of the invention, a method for computing a resultant and a free term of a scaled inverse of a first polynomial v(x) modulo a second polynomial f(x), including: receiving the first polynomial v(x) modulo the second polynomial f(x), where the second polynomial is of a form f(x)=x&#xb1;1, where n=2and k is an integer greater than 0; computing lowest two coefficients of a third polynomial g(z) that is a function of the first polynomial and the second polynomial, where where &#x3c1;, &#x3c1;, . . . , &#x3c1;are roots of the second polynomial f(x) over a field; outputting the lowest coefficient of g(z) as the resultant; and outputting the second lowest coefficient of g(z) divided by n as the free term of the scaled inverse of the first polynomial v(x) modulo the second polynomial f(x).

Fast computation of a single coefficient in an inverse polynomial

Homomorphic evaluations of functions are performed. The functions include operation(s). Variants of key switching and modulus switching are described and are performed prior to or after the operation(s). A key switching transformation converts a ciphertext with respect to a first secret key and a first modulus to a ciphertext with respect to a second secret key and a second modulus. A key switching transformation converts a first version of a ciphertext with respect to a first secret key and with some number r bits of precision to a second version of the selected ciphertext with respect to a second keys and with some other number r' bits of precision. The ciphertexts may be operated on as polynomials represented using evaluation representation, which has benefits for multiplication and automorphism. Further, ciphertexts are associated with an estimate of noise, which is used to determine when to perform modulus switching on the ciphertexts.

Homomorphic evaluation including key switching, modulus switching, and dynamic noise management

Homomorphic evaluation of a function is performed on input ciphertext(s), which were encrypted using a public key of an encryption scheme that also includes multiple secret keys and multiple integer moduli. The homomorphic evaluation of the function includes performing operations(s) on the input ciphertexts. The function includes operation(s) including one or more of addition, multiplication, and automorphism. A key-switching transformation is performed on selected input ciphertext(s), and includes converting a first version of a selected ciphertext with respect to a first of the multiple secret keys and a first modulus to a second version of the selected ciphertext with respect to a second of the multiple secret keys and a second modulus, where the second modulus is an integer factor p times the first modulus, p&#x3e;1. Each of the key switching transformations is performed prior to or after the operation(s) are evaluated. Results of the operation(s) are output.

Growing community of inventors

Elmsford, NY, United States of America

Shai Halevi