The importance of Differential Privacy for a data-driven society
Three months ago Udacity gave me the opportunity to participate in one of its fellowship programs, called the Nanodegree Secure and Private AI Scholarship Challenge Program. At first, I thought it was just another technical challenge to win a scholarship. However, in the course of class I began to understand how important third party data privacy is to a data driven society in a world where data leaks are constantly occurring and companies like Google, Facebook and Samsung are supposedly monitoring you. So if you work with data daily like me or share your data anyway, you need to know how to protect yourself and use the data in an ethical manner.
First, what is Differential Privacy?
According to Cynthia Dwork, “Differential Privacy” refers to a promise, made by a date holder like this: “You will not be affected, adversely or otherwise, allowing your data to be used in any study or analysis, no matter what other studies, data sets, or information sources, are available. ”
In other words, Differential Privacy measures the ability of a data privacy pledge to be maintained without third parties being affected. Thus, projects involving AI, such as machine learning models, should be able to learn patterns without exposing the analyzed object, just generalizing the general case.
Why should your business care about data privacy?
In a data driven world where policies are dictated by the behavior observed in user data, the privacy and security of the data manipulated by your company is not only a matter of respect for democracy, but above all of trust to final user. Thus, laws regulating data use have emerged, such as the General Data Protection Regulation, implemented in Europe in 2018.
In addition, keeping your customer comfortable is a key factor in continuing to use your services and products. Imagine, for example, when you receive unfamiliar emails, web recommendations or unexpected calls, isn’t it annoying and invasive !? If there is no paradigm shift anytime soon, these issues could get worse and the world of big data becomes a big trap, causing a race between user privacy and security companies.
From this point on, the trust generated by your business has a number of benefits, including:
- Compliance Requirements
- Brand Value
- Continuous Growth
- Competitive advantage
So make sure your company uses the data transparently and follows the rules that apply to the places where it operates.
What is the future of our society given the vast amount of data?
Given the hype we are living, data governance has been the main focus of large companies, especially those with products and services linked to continuous human interaction. However, we see that security and privacy on the Internet in general, long since set aside, is taking hold again with the new wave of artificial intelligence. Thus, it will be noted that soon we will be living the hype of privacy and information security, especially regarding the ability to model inference and handle large data sets.
Finally, for AI practitioners, remember to prevent your models from memorizing particular patterns so that third parties cannot reconstruct those patterns and possibly uncover sensitive information.
So, inspired by Deming quotes, I would say that “In data we trust, but give us privacy” — Eduardo Humberto R. da Silva
Get in touch with me: https://www.linkedin.com/in/eduhumb/