2019
DOI: 10.1007/978-3-030-10970-7_21
|View full text |Cite
|
Sign up to set email alerts
|

Unsupervised Machine Learning on Encrypted Data

Abstract: In the context of Fully Homomorphic Encryption, which allows computations on encrypted data, Machine Learning has been one of the most popular applications in the recent past. All of these works, however, have focused on supervised learning, where there is a labeled training set that is used to configure the model. In this work, we take the first step into the realm of unsupervised learning, which is an important area in Machine Learning and has many real-world applications, by addressing the clustering proble… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
49
0

Year Published

2019
2019
2020
2020

Publication Types

Select...
5
3

Relationship

0
8

Authors

Journals

citations
Cited by 37 publications
(49 citation statements)
references
References 36 publications
0
49
0
Order By: Relevance
“…Many recent works focus on clustering in the outsourcing setting (many parties and a trusted/untrusted mediator) [30,37,48,52], or differential privacy setting [8,51,53,54,62]. There are few recent works [18,28,43,58] that consider privacy preserving K-means clustering with full privacy guarantees. The solution of [18] only works for horizontally partitioned data.…”
Section: Related Workmentioning
confidence: 99%
See 2 more Smart Citations
“…Many recent works focus on clustering in the outsourcing setting (many parties and a trusted/untrusted mediator) [30,37,48,52], or differential privacy setting [8,51,53,54,62]. There are few recent works [18,28,43,58] that consider privacy preserving K-means clustering with full privacy guarantees. The solution of [18] only works for horizontally partitioned data.…”
Section: Related Workmentioning
confidence: 99%
“…Moreover, it is not clear how to compute the distance metric in this work. The protocols [28,58] are heavily based on homomorphic encryption and do not scale for large datasets (e.g. more than 10,000 data entries).…”
Section: Related Workmentioning
confidence: 99%
See 1 more Smart Citation
“…Sakellariou and Gounaris [182] perform k-means clustering of encrypted data using Brakerski's FHE scheme [38], but require decryption by a trusted server during analysis. Jäschke and Armknecht present a modiication of k-means that can be evaluated without intermediate decryption, but the estimated run time for a 2-dimensional data set with 400 points is over 25 days on a CPU [129].…”
Section: Clusteringmentioning
confidence: 99%
“…The Paillier cryptosystem is used to encrypt the plaintext data, and then the plaintext operation is replaced with the ciphertext security protocol, but the computational cost is too large. Jaschke and Armknecht [27] solves the division problem in ciphertext operation, and does not allow direct division of two ciphertext, but can divide a ciphertext data by a constant. This constant represents the sum of data, and even if exposed, it will not reveal the key information.…”
Section: A Related Workmentioning
confidence: 99%