Big But Personal Data: How Human Behavior Bounds Privacy and What We Can We Do About It

Big but personal (meta)data

How Human Behavior Bounds Privacy and What We Can We Do About It

Yves-Alexandre de Montjoye @yvesalexandre MIT Media Lab

12 points

Is the way you move around

as unique as your fingerprint

We can use points to identify a fingerprint

Scott

From 10 to 11am

1 km²

1 point for mobility data

~

2 points

Around 11:30am

3 points

For lunch

Boston

How many points do I need to uniquely identify a

mobility traces?

De-identification

Entire country of 1.5 millions people

Our behavior is unique enough

4 points

Identify 95% of people

de Montjoye, Y. A., Hidalgo, C. A., Verleysen, M., & Blondel, V. D. (2013). Unique in the Crowd: The privacy bounds of human mobility. Nature SRep, 3.

What it means

1. It is possible to re-identify mobile phone metadata (even if there is no name or phone number)

Resolution: 800 pixels

Resolution: 300 pixels

Resolution: 150 pixels

Resolution: 75 pixels

Resolution: 30 pixels

Where’s Thierry ?

?

4pm – 10pm 7pm-8pm

Estimating Privacy

Spatial resolution Temporal resolution

Number of points

de Montjoye, Y. A., Hidalgo, C. A., Verleysen, M., & Blondel, V. D. (2013). Unique in the Crowd: The privacy bounds of human mobility. Nature SRep, 3.

Harder to find people

Much easier to find people

Harder to find people

What it means

1. It is possible to re-identify mobile phone metadata (even if there is no name or phone number)

2. It is not simply a question of coarsening the data (we’d just need a few more points)

BFI: Personality test

BFI: Personality test

Behavioral indicators derived from metadata using the Bandicoot toolbox

Predicting personality using metadata

de Montjoye, Y. A., Quoidbach, J., Robic, F., & Pentland, A. S. (2013). Predicting personality using novel mobile phone-based metrics. In Social Computing, Behavioral-Cultural Modeling and Prediction (pp. 48-55). Springer Berlin Heidelberg.

What it means

1. It is possible to re-identify mobile phone metadata (even if there is no name or phone number)

2. It is not simply a question of coarsening the data (we’d just need a few more points)

3. It is not “just” metadata or what is directly visible in the data (e.g. one might use it to predict your personality)

Eagle, N., de Montjoye, Y-A.., & Bettencourt, L. M. (2009). Community computing: Comparisons between rural and urban societies using mobile phone data. IEEE Computational Science and Engineering

We should use this data

Deville, P. et al. (2014). Dynamic population mapping using mobile phone data. Proceedings of the National Academy of Sciences, 201408439.

Wesolowski, A., Eagle, N., Tatem, A. J., Smith, D. L., Noor, A. M., Snow, R. W., & Buckee, C. O. (2012). Quantifying the impact of human mobility on malaria. Science, 338(6104), 267-270.

(but in a privacy-conscientious way)

We should use this data

by: understanding what the real risks are

and designing solutions

Privacy-conscientious anonymization

de Montjoye, Y. A., Smoreda, Z., Trinquart, R., Ziemlicki, C., & Blondel, V. D. (2014). D4D-Senegal: The Second Mobile Phone Data for Development Challenge. arXiv preprint arXiv:1407.4885.

e.g. 2-week mobility traces of 27 x 300.000 individuals + Bandicoot’s behavioral indicators

Online systems: from privacy to security

openPDS/SafeAnswers: - Only shares answers, not raw data - Security mechanisms

openPDS/SafeAnswers

de Montjoye Y.-A., Wang S., Pentland A., On the Trusted Use of Large-Scale Personal Data. IEEE Data Engineering Bulletin, 35-4 (2012). de Montjoye, Y. A., Shmueli, E., Wang, S. S., & Pentland, A. S. (2014). openPDS: Protecting the Privacy of Metadata through SafeAnswers. PLoS ONE, 9(7), e98790.

Yves-Alexandre de Montjoye MIT Media Lab

@yvesalexandre http://deMontjoye.com

In collaboration with Alex “Sandy” Pentland, César Hidalgo, Vincent Blondel, Cameron Kerry, Jake Kendall, Michel Verleysen, Erez Shmueli, Arek Stopczynski, Sune Lehmann, Eaman Jahani