The Experts below are selected from a list of 21 Experts worldwide ranked by ideXlab platform
Takashi Tanaka - One of the best experts on this subject based on the ideXlab platform.
-
Encrypted Value iteration and temporal difference learning over leveled homomorphic encryption
arXiv: Cryptography and Security, 2021Co-Authors: Jihoon Suh, Takashi TanakaAbstract:We consider an architecture of confidential cloud-based control synthesis based on Homomorphic Encryption (HE). Our study is motivated by the recent surge of data-driven control such as deep reinforcement learning, whose heavy computational requirements often necessitate an outsourcing to the third party server. To achieve more flexibility than Partially Homomorphic Encryption (PHE) and less computational overhead than Fully Homomorphic Encryption (FHE), we consider a Reinforcement Learning (RL) architecture over Leveled Homomorphic Encryption (LHE). We first show that the impact of the encryption noise under the Cheon-Kim-Kim-Song (CKKS) encryption scheme on the convergence of the model-based tabular Value Iteration (VI) can be analytically bounded. We also consider secure implementations of TD(0), SARSA(0) and Z-learning algorithms over the CKKS scheme, where we numerically demonstrate that the effects of the encryption noise on these algorithms are also minimal.
Tanaka Takashi - One of the best experts on this subject based on the ideXlab platform.
-
Encrypted Value Iteration and Temporal Difference Learning over Leveled Homomorphic Encryption
2021Co-Authors: Suh Jihoon, Tanaka TakashiAbstract:We consider an architecture of confidential cloud-based control synthesis based on Homomorphic Encryption (HE). Our study is motivated by the recent surge of data-driven control such as deep reinforcement learning, whose heavy computational requirements often necessitate an outsourcing to the third party server. To achieve more flexibility than Partially Homomorphic Encryption (PHE) and less computational overhead than Fully Homomorphic Encryption (FHE), we consider a Reinforcement Learning (RL) architecture over Leveled Homomorphic Encryption (LHE). We first show that the impact of the encryption noise under the Cheon-Kim-Kim-Song (CKKS) encryption scheme on the convergence of the model-based tabular Value Iteration (VI) can be analytically bounded. We also consider secure implementations of TD(0), SARSA(0) and Z-learning algorithms over the CKKS scheme, where we numerically demonstrate that the effects of the encryption noise on these algorithms are also minimal.Comment: 8 pages, 7 figures, American Control Conferenc
Jihoon Suh - One of the best experts on this subject based on the ideXlab platform.
-
Encrypted Value iteration and temporal difference learning over leveled homomorphic encryption
arXiv: Cryptography and Security, 2021Co-Authors: Jihoon Suh, Takashi TanakaAbstract:We consider an architecture of confidential cloud-based control synthesis based on Homomorphic Encryption (HE). Our study is motivated by the recent surge of data-driven control such as deep reinforcement learning, whose heavy computational requirements often necessitate an outsourcing to the third party server. To achieve more flexibility than Partially Homomorphic Encryption (PHE) and less computational overhead than Fully Homomorphic Encryption (FHE), we consider a Reinforcement Learning (RL) architecture over Leveled Homomorphic Encryption (LHE). We first show that the impact of the encryption noise under the Cheon-Kim-Kim-Song (CKKS) encryption scheme on the convergence of the model-based tabular Value Iteration (VI) can be analytically bounded. We also consider secure implementations of TD(0), SARSA(0) and Z-learning algorithms over the CKKS scheme, where we numerically demonstrate that the effects of the encryption noise on these algorithms are also minimal.
Suh Jihoon - One of the best experts on this subject based on the ideXlab platform.
-
Encrypted Value Iteration and Temporal Difference Learning over Leveled Homomorphic Encryption
2021Co-Authors: Suh Jihoon, Tanaka TakashiAbstract:We consider an architecture of confidential cloud-based control synthesis based on Homomorphic Encryption (HE). Our study is motivated by the recent surge of data-driven control such as deep reinforcement learning, whose heavy computational requirements often necessitate an outsourcing to the third party server. To achieve more flexibility than Partially Homomorphic Encryption (PHE) and less computational overhead than Fully Homomorphic Encryption (FHE), we consider a Reinforcement Learning (RL) architecture over Leveled Homomorphic Encryption (LHE). We first show that the impact of the encryption noise under the Cheon-Kim-Kim-Song (CKKS) encryption scheme on the convergence of the model-based tabular Value Iteration (VI) can be analytically bounded. We also consider secure implementations of TD(0), SARSA(0) and Z-learning algorithms over the CKKS scheme, where we numerically demonstrate that the effects of the encryption noise on these algorithms are also minimal.Comment: 8 pages, 7 figures, American Control Conferenc
Maroti Deshmukh - One of the best experts on this subject based on the ideXlab platform.
-
image compression and encryption using chinese remainder theorem
Multimedia Tools and Applications, 2019Co-Authors: Tejas Duseja, Maroti DeshmukhAbstract:The proposed encryption technique uses Chinese Remainder Theorem (CRT) and hash map to generate and distribute secret co-prime keys to participants. It utilizes the fact that CRT gives a unique solution for the set of congruent equations if and only if the modulus Values are relatively co-prime to each other i.e., Greatest Common Divisor (GCD) of moduli is equal to 1. In this paper, we have proposed secret image sharing scheme using CRT and obtained results on grayscale images of different dimensions. The proposed technique not only increases randomness in Encrypted image but also compresses the image, resulting in easy storage and fast transmission. As compression ratio is dependent on shared keys, all shared keys are essential for recovering image without any noise, absence of any key gives an image which is deviated from our original image. For r participants, r pixels are Encrypted and compressed simultaneously at a time using CRT which gives one Encrypted unique Value corresponding to those r pixels. As this Encrypted Value can be greater than 255, hash map is used to store this Value. The experimental results show that the Encrypted image is compressed, is not disclosing any secret information and recovery of original image is loss-less.