# Overview

The files:

    sha1sum                                    filename
    3fddee5583ceef3e6b1b314e2fdadfbc4fe733d5   cross_validation.m
    663810422e475a0f7e2ba96af94560b6d3da8ced   training_opt.m
    3d85d65d2bdb114bbd7725108a7636024834b5a0   testing_opt.m

as referenced in the following paper:

    @inproceedings{Sung:2014,
        Author = { Keen Sung and Brian Neil Levine and Marc Liberatore},
        Booktitle = {Proc. IEEE Workshop on Mobile System Technologies (MoST)},
        Keywords = {Privacy; Cellular; Remote Inference; security},
        Month = {May},
        Sponsors = {CNS-0905349},
        Title = {{Location Privacy without Carrier Cooperation}},
        Url = {http://forensics.umass.edu/pubs/Sung-MoST-2014.pdf},
        Year = {2014}
    }

contain the code for the passive attack experiments. It requires the Reality Mining
dataset to be loaded into MATLAB.

cross_validation.m trains on single months of data and tests on the respective following months.
In each case, it tests different lengths of data (1 minute, 1 hour, 1 day, 1 week, 1 month) and
different update strategies (always-update, forming LA, LA only) using the functions training_opt.m
and testing_opt.m.

# Usage

1. Download the Reality Mining dataset from http://realitycommons.media.mit.edu/realitymining.html

2. Load realitymining.mat into MATLAB

3. Run cross_validation.m

# History

First written by Keen Sung <ksung@cs.umass.edu> on 2014-05-17