Muutke küpsiste eelistusi

Privacy-Preserving Machine Learning [Pehme köide]

4.20/5 (10 hinnangut Goodreads-ist)
, ,
  • Formaat: Paperback / softback, 300 pages, kõrgus x laius x paksus: 234x186x18 mm, kaal: 620 g
  • Ilmumisaeg: 21-Apr-2023
  • Kirjastus: Manning Publications
  • ISBN-10: 1617298042
  • ISBN-13: 9781617298042
Teised raamatud teemal:
Privacy-Preserving Machine LearningSuurem pilt
  • Formaat: Paperback / softback, 300 pages, kõrgus x laius x paksus: 234x186x18 mm, kaal: 620 g
  • Ilmumisaeg: 21-Apr-2023
  • Kirjastus: Manning Publications
  • ISBN-10: 1617298042
  • ISBN-13: 9781617298042
Teised raamatud teemal:
Püsilink: https://www.kriso.ee/db/9781617298042.html
Märksõnad:
Privacy-Preserving Machine Learning is a practical guide to keeping ML data anonymous and secure. You'll learn the core principles behind different privacy preservation technologies, and how to put theory into practice for your own machine learning.

Complex privacy-enhancing technologies are demystified through real world use cases forfacial recognition, cloud data storage, and more. Alongside skills for technical implementation, you'll learn about current and future machine learning privacy challenges and how to adapt technologies to your specific needs. By the time you're done, you'll be able to create machine learning systems that preserve user privacy without sacrificing data quality and model performance.

Large-scale scandals such as the Facebook Cambridge Analytic a data breach have made many users wary of sharing sensitive and personal information. Demand has surged among machine learning engineers for privacy-preserving techniques that can keep users private details secure without adversely affecting the performance of models.

Arvustused

An interesting and well-structured book about an emerging discipline that will certainly keep growing in importance. Alain Couniot

Gives a deep and thorough introduction into preserving privacy while using personal data for machine learning and data mining. HaraldKuhn

Makes for a great introduction to privacy-preserving techniques whichmake full use of machine learning. Aditya Kaushik

An interesting book under a rising hot topic: privacy. I like the way using examples and figures to illustrate concepts. XiangboMao

The only book, that I am aware of, which goes into depths of data privacy while making sure it doesn't get boring for readers. VishweshRavi Shrimali



An interesting book under a rising hot topic: privacy. XiangboMao

A great resource to understand privacy preserving ML. DhivyaSivasubramanian

A great book to getting a deep theoretical overview of the landscape of privacy preserving approaches while also getting hands-on pragmatic experience.Stephen Oates

Preface xii
Acknowledgments xiv
About this book xv
About the authors xviii
About the cover illustration xx
Part 1 Basics of privacy-preserving machine learning with differential privacy
1(92)
1 Derations in machine learning
3(22)
1.1 Privacy complications in the AI era
4(1)
1.2 The threat of learning beyond the intended purpose
5(3)
Use of private data on the fly
6(1)
How data is processed inside ML algorithms
6(1)
Why privacy protection in ML is important
7(1)
Regulatory requirements and the utility vs. privacy tradeoff
7(1)
1.3 Threats and attacks for ML systems
8(8)
The problem of private data in the clear
9(1)
Reconstruction attacks
9(3)
Model inversion attacks
12(1)
Membership inference attacks
13(2)
De-anonymization or re-identification attacks
15(1)
Challenges of privacy protection in big data analytics
15(1)
1.4 Securing privacy while learning from data: Privacy-preserving machine learning
16(6)
Use of differential privacy
17(1)
Local differential privacy
18(1)
Privacy-preserving synthetic data generation
18(1)
Privacy-preserving data mining techniques
19(2)
Compressive privacy
21(1)
1.5 How is this book structured?
22(3)
2 Differential privacy for machine learning
25(31)
2.1 What is differential privacy?
26(9)
The concept of differential privacy
27(3)
How differential privacy works
30(5)
2.2 Mechanisms of differential privacy
35(13)
Binary mechanism (randomized response)
35(3)
Laplace mechanism
38(5)
Exponential mechanism
43(5)
2.3 Properties of differential privacy
48(8)
Postprocessing property of differential privacy
48(2)
Group privacy property of differential privacy
50(1)
Composition properties of differential privacy
51(5)
3 Advanced concepts of differential privacy for machine learning
56(37)
3.1 Applying differential privacy in machine learning
57(5)
Input perturbation
58(1)
Algorithm perturbation
59(1)
Output perturbation
60(1)
Objective perturbation
60(2)
3.2 Differentially private supervised learning algorithms
62(12)
Differentially private naive Bayes classification
62(6)
Differentially private logistic regression
68(4)
Differentially private linear regression
72(2)
3.3 Differentially private unsupervised learning algorithms
74(3)
Differentially private k-means clustering
74(3)
3.4 Case study: Differentially private principal component analysis
77(16)
The privacy of PCA over horizontally partitioned data
78(2)
Designing differentially private PCA over horizontally partitioned data
80(5)
Experimentally evaluating the performance of the protocol
85(8)
Part 2 Local differential privacy and synthetic data generation
93(84)
4 Local differential privacy for machine learning
95(28)
4.1 What is local differential privacy?
96(8)
The concept of local differential privacy
97(4)
Randomized response for local differential privacy
101(3)
4.2 The mechanisms of local differential privacy
104(19)
Direct encoding
104(6)
Histogram encoding
110(7)
Unary encoding
117(6)
5 Advanced LDP mechanisms for machine learning
123(23)
5.1 A quick recap of local differential privacy
124(1)
5.2 Advanced LDP mechanisms
125(6)
The Laplace mechanism for LDP
126(1)
Duchi's mechanism for LDP
127(2)
The Piecewise mechanism for LDP
129(2)
5.3 A case study implementing LDP naive Bayes classification
131(15)
Using naive Bayes with ML classification
131(1)
Using LDP naive Bayes with discrete features
132(5)
Using LDP naive Bayes with continuous features
137(3)
Evaluating the performance of different LDP protocols
140(6)
6 Privacy-preserving synthetic data generation
146(31)
6.1 Overview of synthetic data generation
148(3)
What is synthetic data? Why is it important?
148(1)
Application aspects of using synthetic data for privacy preservation
149(1)
Generating synthetic data
150(1)
6.2 Assuring privacy via data anonymization
151(4)
Private information sharing vs. privacy concerns
151(1)
Using k-anonymity against re-identification attacks
152(2)
Anonymization beyond k-anonymity
154(1)
6.3 DP for privacy-preserving synthetic data generation
155(13)
DP synthetic histogram representation generation
156(4)
DP synthetic tabular data generation
160(2)
DP synthetic multi-marginal data generation
162(6)
6.4 Case study on private synthetic data release via feature4evel micro-aggregation
168(9)
Using hierarchical clustering and micro-aggregation
168(1)
Generating synthetic data
169(2)
Evaluating the performance of the generated synthetic data
171(6)
Part 3 Building privacy-assured machine learning applications
177(114)
7 Privacy-preserving data mining techniques
179(23)
7.1 The importance of privacy preservation in data mining and management
180(3)
7.2 Privacy protection in data processing and mining
183(2)
What is data mining and how is it used?
183(1)
Consequences of privacy regulatory requirements
184(1)
7.3 Protecting privacy by modifying the input
185(1)
Applications and limitations
186(1)
7.4 Protecting privacy when publishing data
186(16)
Implementing data sanitization operations in Python
189(4)
k-anonymity
193(5)
Implementing k-anonymity in Python
198(4)
8 Privacy-preserving data management and operations
202(31)
8.1 A quick recap of privacy protection in data processing and mining
203(1)
8.2 Privacy protection beyond k-anonymity
204(12)
t-diversity
205(3)
t-closeness
208(3)
Implementing privacy models with Python
211(5)
8.3 Protecting privacy by modifying the data mining output
216(4)
Association rule hiding
217(1)
Reducing the accuracy of data mining operations
218(1)
Inference control in statistical databases
219(1)
8.4 Privacy protection in data management systems
220(13)
Database security and privacy: Threats and vulnerabilities
221(1)
How likely is a modern database system to leak private information?
222(1)
Attacks on database systems
222(3)
Privacy-preserving techniques in statistical database systems
225(3)
What to consider when designing a customizable privacy-preserving database system
228(5)
9 Compressive privacy for machine learning
233(35)
9.1 Introducing compressive privacy
235(2)
9.2 The mechanisms of compressive privacy
237(2)
Principal component analysis (PCA)
237(1)
Other dimensionality reduction methods
238(1)
9.3 Using compressive privacy for ML applications
239(12)
Implementing compressive privacy
240(6)
The accuracy of the utility task
246(3)
The effect of p' in DCA for privacy and utility
249(2)
9.4 Case study: Privacy-preserving PCA and DCA on horizontally partitioned data
251(17)
Achieving privacy preservation on horizontally partitioned data
253(1)
Recapping dimensionality reduction approaches
254(1)
Using additive homomorphic encryption
255(1)
Overview of the proposed approach
256(2)
How privacy-preserving computation works
258(5)
Evaluating the efficiency and accuracy of the privacy-preserving PCA and DCA
263(5)
10 Putting it all together: Designing a privacy-enhanced platform (DataHub)
268(23)
10.1 The significance of a research data protection and sharing platform
270(2)
The motivation behind the DataHub platform
270(1)
DataHub's important features
271(1)
10.2 Understanding the research collaboration workspace
272(8)
The architectural design
275(1)
Blending different trust models
276(2)
Configuring access control mechanisms
278(2)
10.3 Integrating privacy and security technologies into DataHub
280(11)
Data storage with a cloud-based secure NoSQL database
280(2)
Privacy preserving data collection with local differential privacy
282(2)
Privacy-preserving machine learning
284(2)
Privacy preserving query processing
286(1)
Using synthetic data generation in the DataHub platform
287(4)
Appendix: More details about differential privacy 291(8)
References 299(6)
Index 305
J. Morris Chang is a professor in the Department of Electrical Engineering of University of South Florida, Tampa, USA. He received his PhDfrom North Carolina State University. Since 2012, his research projects on cybersecurity and machine learning have been funded by DARPA and agencies under DoD. He hasled a DARPA project under the Brandeis Program, focusing on privacy-preserving computation over the internet for three years.

Di Zhuang received his BSc degree in computer science and information security from Nankai University, Tianjin, China. He is currently a PhD candidate in the Department of Electrical Engineering of University of South Florida, Tampa, USA. Heconducted privacy-preserving machine learning research under the DARPA Brandeis Program from 2015 to 2018.

G. Dumindu Samaraweera received his BSc degree in computer systems and networking from Curtin University, Australia, and a MSc in enterprise application development degree from Sheffield Hallam University, UK. He is currently reading for his PhD in electrical engineering at University of South Florida, Tampa.