公开(公告)号：US11575500B2

公开(公告)日：2023-02-07

申请号：US16045113

申请日：2018-07-25

Applicant: SAP SE

Inventor： Laurent Y. Gomez ,  Jose Marquez ,  Patrick Duverger

IPC: H04L9/00 ,  G06N3/04 ,  G06N3/08 ,  H04L9/08

Abstract: Systems and methods are provided for receiving input data to be processed by an encrypted neural network (NN) model, and encrypting the input data using a fully homomorphic encryption (FHE) public key associated with the encrypted NN model to generate encrypted input data. The systems and methods further provided for processing the encrypted input data to generate an encrypted inference output, using the encrypted NN model by, for each layer of a plurality of layers of the encrypted NN model, computing an encrypted weighted sum using encrypted parameters and a previous encrypted layer, the encrypted parameters comprising at least an encrypted weight and an encrypted bias, approximating an activation function for the level into a polynomial, and computing the approximated activation function on the encrypted weighted sum to generate an encrypted layer. The generated encrypted inference output is sent to a server system for decryption.

公开(公告)号：US09083750B2

公开(公告)日：2015-07-14

申请号：US13941882

申请日：2013-07-15

Applicant: SAP SE

Inventor： Laurent Y. Gomez ,  Ivonne Scherfenberg

IPC: H04L29/06

CPC classification number: H04L63/20 ,  H04L63/08 ,  H04L63/104 ,  H04L63/105

Abstract: A computer-implemented method for authentication involves defining a level of trust required for access to a resource independently of any particular authentication mechanism or instance, determining levels of trust associated with a plurality of authentication instances, and selecting and combining two or more of the authentication instances to meet or exceed the required level of trust.

Abstract translation: 用于认证的计算机实现的方法包括定义独立于任何特定认证机制或实例的访问资源所需的信任级别，确定与多个认证实例相关联的信任级别，以及选择和组合两个或多个认证 满足或超过所要求的信任水平的实例。

公开(公告)号：US12047393B2

公开(公告)日：2024-07-23

申请号：US17401873

申请日：2021-08-13

Applicant: SAP SE

Inventor： Laurent Y. Gomez ,  Cedric R. J. Hebert ,  Slim Trabelsi

IPC: H04L29/06 ,  G06N20/20 ,  H04L9/40

CPC classification number: H04L63/1416 ,  G06N20/20 ,  H04L63/1425 ,  H04L63/20

Abstract: In an example embodiment, a combination of machine learning and rule-based techniques are used to automatically detect social engineering attacks in a computer system. More particularly, three phases of detection are utilized on communications in a thread or stream of communications: attack contextualization, intention classification, and security policy violation detection. Each phase of detection causes a score to be generated that is reflective of the degree of danger in the thread or stream of communications, and these scores may then be combined into a single global social engineering attack score, which then may be used to determined appropriate actions to deal with the attack if it transgresses a threshold.

公开(公告)号：US20200036510A1

公开(公告)日：2020-01-30

申请号：US16045113

申请日：2018-07-25

Applicant: SAP SE

Inventor： Laurent Y. Gomez ,  Jose Marquez ,  Patrick Duverger

IPC: H04L9/00 ,  H04L9/08 ,  G06N3/08 ,  G06N3/04

Abstract: Systems and methods are provided for receiving input data to be processed by an encrypted neural network (NN) model, and encrypting the input data using a fully homomorphic encryption (FHE) public key associated with the encrypted NN model to generate encrypted input data. The systems and methods further provided for processing the encrypted input data to generate an encrypted inference output, using the encrypted NN model by, for each layer of a plurality of layers of the encrypted NN model, computing an encrypted weighted sum using encrypted parameters and a previous encrypted layer, the encrypted parameters comprising at least an encrypted weight and an encrypted bias, approximating an activation function for the level into a polynomial, and computing the approximated activation function on the encrypted weighted sum to generate an encrypted layer. The generated encrypted inference output is sent to a server system for decryption.

公开(公告)号：US20240396905A1

公开(公告)日：2024-11-28

申请号：US18763984

申请日：2024-07-03

Applicant: SAP SE

Inventor： Laurent Y. Gomez ,  Cedric R.J. Hebert ,  Slim Trabelsi

IPC: H04L9/40 ,  G06N20/20

Abstract: In an example embodiment, a combination of machine learning and rule-based techniques are used to automatically detect social engineering attacks in a computer system. More particularly, three phases of detection are utilized on communications in a thread or stream of communications: attack contextualization, intention classification, and security policy violation detection. Each phase of detection causes a score to be generated that is reflective of the degree of danger in the thread or stream of communications, and these scores may then be combined into a single global social engineering attack score, which then may be used to determined appropriate actions to deal with the attack if it transgresses a threshold.

公开(公告)号：US11816546B2

公开(公告)日：2023-11-14

申请号：US17983751

申请日：2022-11-09

Applicant: SAP SE

Inventor： Laurent Y. Gomez

IPC: G06F21/60 ,  G06N20/00 ,  G06N3/02

CPC classification number: G06N20/00 ,  G06F21/602 ,  G06N3/02

Abstract: Fairness and output authenticity for secure distributed machine learning is provided by way of an encrypted output of a garbled circuit which is simultaneously provided to a garbler and an evaluator by an output discloser. Related systems, methods and articles of manufacture are also disclosed.

公开(公告)号：US20230046392A1

公开(公告)日：2023-02-16

申请号：US17401873

申请日：2021-08-13

Applicant: SAP SE

Inventor： Laurent Y. Gomez ,  Cedric R.J. Hebert ,  Slim Trabelsi

IPC: H04L29/06 ,  G06N20/20

Abstract: In an example embodiment, a combination of machine learning and rule-based techniques are used to automatically detect social engineering attacks in a computer system. More particularly, three phases of detection are utilized on communications in a thread or stream of communications: attack contextualization, intention classification, and security policy violation detection. Each phase of detection causes a score to be generated that is reflective of the degree of danger in the thread or stream of communications, and these scores may then be combined into a single global social engineering attack score, which then may be used to determined appropriate actions to deal with the attack if it transgresses a threshold.