Section 3 Applications

measures for social networks under active attack. Information Sciences.

*Security and Privacy From a Legal, Ethical, and Technical Perspective*

[19] Johnson DS. Approximation algorithms for combinatorial problems. Journal of Computer and System Sciences. 1974;**9**(3):256-278

[20] Zachary WW. An information flow model for conflict and fission in small groups. Journal of Anthropological Research. 1977;**33**(4):452-473

[21] Loomis CP, Morales JO, Clifford RA, Leonard OE. Turrialba: Social Systems and the Introduction of Change. Glencoe, IL: Free Press; 1953

[22] Gleiser PM, Danon L. Community structure in jazz. Advances in Complex

[23] Guimera R, Danon L, Diaz-Guilera A, Giralt F, Arenas A. Self-similar community structure in a network of human interactions. Physical Review E.

[24] Enron email network. Available from: UC Berkeley Enron Email Analysis website http://bailando.sims. berkeley.edu/enron\_email.html

[25] Paranjape A, Benson AR, Leskovec J.

[26] Panzarasa P, Opsahl T, Carley KM. Patterns and dynamics of users' behavior and interaction: Network analysis of an online community. Journal of the American Society for Information Science and Technology.

[27] Hamsterster friendships network dataset–KONECT, April 2017. Available from: http://konect.uni-koblenz.de/ networks/petster-friendships-hamster)

Motifs in temporal networks. In: Proceedings of the 10th ACM International Conference on Web Search and Data Mining; ACM. 2017.

Systems. 2003;**6**(04):565-573

2003;**68**(6):065103

pp. 601-610

2009;**60**(5):911-932

**138**

2019;**473**:87-100

**141**

**Chapter 9**

**Abstract**

Networks

in both academia and industry.

comes with significant limitation (e.g., [1]).

**1. Introduction**

following:

that mechanism.

Beyond Differential Privacy:

with Deep Generative Neural

*Ofer Mendelevitch and Michael D. Lesh*

Synthetic Micro-Data Generation

Recent advances in generative modeling, based on large scale deep neural networks, provide a novel approach for sharing individual-level datasets (micro-data) without privacy concerns. Unlike differential privacy, which enforces a specific query mechanism on data to ensure privacy, generative models can accurately learn the statistical patterns of such micro-data and then be used to generate "synthetic data" that accurately reflects these statistical patterns, yet contain none of the original data itself, and thus can be safely shared for analysis and modeling without compromising privacy. The successful application of these techniques to various industries including healthcare, finance, and autonomous vehicles is promising and results in continued investment in research and development of generative models

**Keywords:** generative models, synthetic data, deep neural networks, micro-data

Differential privacy, created more than a decade ago, continues to play an important role in protecting privacy of micro-data while enabling statistical analysis. Initially applied by statistics agencies such as the US census bureau, it is now well recognized that, although useful for some applications, differential privacy

To understand some of the limitations of differential privacy, consider the

• Differential privacy is defined around the concept of a *mechanism*; as such, it is not intended to create "sharable datasets," but instead allows a user (analyst) to submit various types of queries (via the defined *query mechanism*), requesting some kind of aggregate statistics, like summary statistics of the original data. This limits the usability of differential privacy to queries that are supported by

• An appropriate *privacy budget* needs to be decided upon, and in practice it's often difficult to agree on what that budget needs to be. In fact, practical
