Developers
July 31, 2020

Facebook Protects Privacy Mobility Data During the Pandemic

Two metrics help inform researchers and public health experts about how populations respond to social distancing. Change in Movement and Stay Put.
Source: Unsplash

Today we will talk about a Facebook release. They are releasing the Facebook Movement Range Maps. Facebook has also released three additions to its Disease Prevention Maps product. This product helps health researchers and organizations to respond effectively to Covid-19.

The organizations that are using the product are Harvard Chan School of Public Health, The National Tsing Hua University, The University of Venice, the Direct Relief, The Bill and Melinda Gates Foundation, and The World Bank.

To preserve the privacy and make it a priority, Facebook applies a new differential privacy (DP) framework. It minimizes the risk of reidentification of individual data. The focus of using a DP framework is that nobody can reidentify users.

Facebook Movement Range Data

Two metrics help the data sets inform researchers and public health experts about how populations respond to social distancing measures. The metrics are, Change in Movement and Stay Put.

Change in Movement focuses on how much people are moving around and then compares the data with a baseline of social distancing measures. Stay Put, on the other hand, looks at the fraction of the population that appears to stay within a small area in one single day.  

Movement Range Trends are produced by the use of data of Facebook's precise location like nearby friends and finding wi-fi to et local content and ads. Only people that opt into Location are included in the service.

To generate data for each region where the service works, the location is aggregated to the users that are spending evenings at that location. First, the region is mapped, and then the service ensures there is enough data to produce meaningful trends to protect the privacy of the users. The minimum qualifying number is 300 people, if there is less, then the region is omitted from the data sets.

Change in Movement shows the decrease of people moving around since the onset of the coronavirus pandemic. How many people move around is counted by the number of level-16 Bing tiles. The level-16 Bing tiles are 600 meters by 600 meters area sizes.  

The more people are seen in more tiles, the more they are moving around. All the eligible people in a given region are taken each day and compute the number of distinct tiles that they were seen at.  

To prevent high volumes of active people from affecting the data and to limit the amount of noise, differential privacy is added. The maximum value is equal to 200 tiles. Even if a user is contributing more than 200 tiles, the system will only count 200 tiles per user. The result of all the values summed, gets the total number of tiles for each region.

Differential privacy framework

A differential privacy framework is employed to protect the privacy and provide a mathematical limit on the risk that an individual can take from the data. How do they do this? An appropriate amount of noise is added to the total files visited per region.

The amount of noise is proportional to the sensitivity of the data. In the total tiles visited, the sensitivity equals the most tiles visited by the individual, as we have previously seen, with a maximum size of 200 tiles.  

The calculation of differential privacy noise counts with an important value known as epsilon. This parameter is meant to control the level of privacy protection reached by the addition of noise.

Lastly, the noise is generated by drawing a Laplace distribution. The distribution takes two parameters, a location and a diversity parameter. The total number of people in each region is divided to get the noisy average number of tiles visited.  

Based on this, the noisy average number of tiles visited per region can be computed for any given day. For the majority of places in the world, Facebook uses four weeks of February as the considered period.

The Stay Put metric calculates how many people were staying near or at home. It calculates it by calculating the percentage of eligible people who are observed in a single level-16 Bing tile in one day.

In conclusion, Facebook is taking seriously the analysis of data and privacy protection during the pandemic. Facebook has released the Movement Range Maps. It is used by the most important organizations worldwide. Including Harvard Chan School of Public Health, The National Tsing Hua University, The University of Venice, the Direct Relief, The Bill and Melinda Gates Foundation, and The World Bank. Two metrics help the data sets inform researchers and public health experts about how populations respond to social distancing measures. The metrics are Change in Movement and Stay Put. Change in Movement focuses on how much people are moving around and then compares the data with a baseline of social distancing measures. The Stay Put metric calculates how many people were staying near or at home.

TagsFacebookPandemicData Privacy
Lucas Bonder
Technical Writer
Lucas is an Entrepreneur, Web Developer, and Article Writer about Technology.

Related Articles

Back
DevelopersJuly 31, 2020
Facebook Protects Privacy Mobility Data During the Pandemic
Two metrics help inform researchers and public health experts about how populations respond to social distancing. Change in Movement and Stay Put.

Today we will talk about a Facebook release. They are releasing the Facebook Movement Range Maps. Facebook has also released three additions to its Disease Prevention Maps product. This product helps health researchers and organizations to respond effectively to Covid-19.

The organizations that are using the product are Harvard Chan School of Public Health, The National Tsing Hua University, The University of Venice, the Direct Relief, The Bill and Melinda Gates Foundation, and The World Bank.

To preserve the privacy and make it a priority, Facebook applies a new differential privacy (DP) framework. It minimizes the risk of reidentification of individual data. The focus of using a DP framework is that nobody can reidentify users.

Facebook Movement Range Data

Two metrics help the data sets inform researchers and public health experts about how populations respond to social distancing measures. The metrics are, Change in Movement and Stay Put.

Change in Movement focuses on how much people are moving around and then compares the data with a baseline of social distancing measures. Stay Put, on the other hand, looks at the fraction of the population that appears to stay within a small area in one single day.  

Movement Range Trends are produced by the use of data of Facebook's precise location like nearby friends and finding wi-fi to et local content and ads. Only people that opt into Location are included in the service.

To generate data for each region where the service works, the location is aggregated to the users that are spending evenings at that location. First, the region is mapped, and then the service ensures there is enough data to produce meaningful trends to protect the privacy of the users. The minimum qualifying number is 300 people, if there is less, then the region is omitted from the data sets.

Change in Movement shows the decrease of people moving around since the onset of the coronavirus pandemic. How many people move around is counted by the number of level-16 Bing tiles. The level-16 Bing tiles are 600 meters by 600 meters area sizes.  

The more people are seen in more tiles, the more they are moving around. All the eligible people in a given region are taken each day and compute the number of distinct tiles that they were seen at.  

To prevent high volumes of active people from affecting the data and to limit the amount of noise, differential privacy is added. The maximum value is equal to 200 tiles. Even if a user is contributing more than 200 tiles, the system will only count 200 tiles per user. The result of all the values summed, gets the total number of tiles for each region.

Differential privacy framework

A differential privacy framework is employed to protect the privacy and provide a mathematical limit on the risk that an individual can take from the data. How do they do this? An appropriate amount of noise is added to the total files visited per region.

The amount of noise is proportional to the sensitivity of the data. In the total tiles visited, the sensitivity equals the most tiles visited by the individual, as we have previously seen, with a maximum size of 200 tiles.  

The calculation of differential privacy noise counts with an important value known as epsilon. This parameter is meant to control the level of privacy protection reached by the addition of noise.

Lastly, the noise is generated by drawing a Laplace distribution. The distribution takes two parameters, a location and a diversity parameter. The total number of people in each region is divided to get the noisy average number of tiles visited.  

Based on this, the noisy average number of tiles visited per region can be computed for any given day. For the majority of places in the world, Facebook uses four weeks of February as the considered period.

The Stay Put metric calculates how many people were staying near or at home. It calculates it by calculating the percentage of eligible people who are observed in a single level-16 Bing tile in one day.

In conclusion, Facebook is taking seriously the analysis of data and privacy protection during the pandemic. Facebook has released the Movement Range Maps. It is used by the most important organizations worldwide. Including Harvard Chan School of Public Health, The National Tsing Hua University, The University of Venice, the Direct Relief, The Bill and Melinda Gates Foundation, and The World Bank. Two metrics help the data sets inform researchers and public health experts about how populations respond to social distancing measures. The metrics are Change in Movement and Stay Put. Change in Movement focuses on how much people are moving around and then compares the data with a baseline of social distancing measures. The Stay Put metric calculates how many people were staying near or at home.

Facebook
Pandemic
Data Privacy
About the author
Lucas Bonder -Technical Writer
Lucas is an Entrepreneur, Web Developer, and Article Writer about Technology.

Related Articles