walk2friends

This repository provides a reference implementation of walk2friends as described in the paper:

walk2friends: Inferring Social Links from Mobility Profiles.
Michael Backes, Mathias Humbert, Jun Pang, and Yang Zhang.
The 24th ACM SIGSAC Conference on Computer and Communications Security (CCS).

Basic Usage

To run the code, please go to folder src/
cd src/

Attack

Running our social link inference attack on the New York data with each user having at least 20 check-ins:
python main_attack.py ny 20

Defense

Hiding 60% of the check-ins for defense:
python main_hiding.py ny 20 60

Replacing 60% of the check-ins with a 15 step random walk for defense:
python main_replace.py ny 20 60 15

Utility

Measuring the utility after hiding 60% of the check-ins:
python main_utility_hiding.py ny 20 60

Measuring the utility after replacing 60% of the check-ins with a 15 step random walk:
python main_utility_replace.py ny 20 60 15

Requirements

• pandas
• numpy
• scipy
• scikit-learn

It is recommended to install Anaconda, a python data science distribution, which includes all the above packages.

• gensim
• joblib

Citing

If you find walk2friends useful in your research, please cite the following paper:

@inproceedings{BHPZ17,
author = {Michael Backes and Mathias Humbert and Jun Pang and Yang Zhang},
title = {walk2friends: Inferring Social Links from Mobility Profiles.},
booktitle = {Proceedings of the 24th ACM SIGSAC Conference on Computer and Communications Security (CCS)},
year = {2017},
pages = {1943-1957},
publisher = {ACM}
}

Miscellaneous

If you have any questions about the code and/or the algorithm, please send an email to yang.zhang@cispa.saarland.