- A novel frequency-based differential privacy model for trajectory anonymization
- The global and local mechanisms can be performed individually or composed with different ordering, which can be specified in the parameter file (detailed later).
- An efficient hierarchical grid index is provided along with the bottom-up-down search strategy to support large-scale data.
CMakeLists.txt -- the version of cmake might be revised accordingly (currently is 3.15)
cmake-build/ -- If needed, clean some files, then re-compile and re-build in your device
config.properties -- the program will read parameters here
Testing/ -- all testing data and some possbile outputs
RoadNetworkInfo/NNid2lnglat.csv -- the road network information used for simple map-matching as the preprosseing of data
TestData-tdrive/ -- T-Drive dataset, including 600 objects' one-week data,
note that some objects could be discarded due to few data points
outputs/ -- log file and the anonymized data will be outputed here
...
main.cpp -- the entry of the program
spatial/ -- some basic geometry used in this program
grids/ -- the basic grid defined here (not the hierarchical grid)
io/ -- read/write file
linking/ -- to extract trajectory signatures
privacy/ -- all privacy models and relevant structures defined here (including hierarachical grid)
local/ -- local mechanism implemented here
global/ -- global mechanism implemented here
... -- other paticipant classes and headers (not detailed here)
To compile this project using cmake and c++ compiler, run this script:
./buildCPP.sh
If a permission denied error happens, please try to modify the permission of this script file by chmod 700 buildCPP.sh
To run the program compiled before, execute the command in the folder of cmake-build:
./FrequencyDP
This command will execute the program in main.cpp.
Datasets should be located in the cmake-build/Testing/TestData-tdrive/ folder. To anonymize a new dataset by yourself, simply add it in this folder and modify the parameter in the file config.properties.
The test configuration is in cmake-build/config.properties.
Tested with CentOS Linux (with gcc version 9.2.0) and macOS Monterey (Apple clang version 13.0.0)
RoadNetworkFile = ./Testing/RoadNetworkInfo/NNid2lnglat.csv -- fixed, used for the simple map-matching initially
data_FolderInput = ./Testing/TestData-tdrive/ -- the input data to be protected
data_FolderOutput = ./Testing/outputs/ -- the output folder, any log file or anonymized trajectries will be here
data_total = 100 -- how many objects to be protected (e.g., 100, 500, etc)
output_results = false -- whether the anonymized data will be outputted
linking_reduction = 10 -- for signature reduction, i.e., the number of signature points for each trajectory
privacy_epsilon = 0.5,2.0 -- privacy budget for differential privacy; allow multiple choices here
edit_global = true -- pureGlobal
edit_local = true -- pureLocal
global_local_composition_order = GLLG -- GL: Global->Local; LG: Local->Global; exchangable composition
If you find our algorithms or the experimental results useful, please kindly cite the following paper:
@INPROCEEDINGS{jin2022,
author={Jin, Fengmei and Hua, Wen and Ruan, Boyu and Zhou, Xiaofang},
booktitle={Proceedings of the 38th IEEE International Conference on Data Engineering (ICDE)},
title={Frequency-based Randomization for Guaranteeing Differential Privacy in Spatial Trajectories},
year={2022},
pages={1727-1739},
doi={10.1109/ICDE53745.2022.00175}
}
Please feel free to contact fengmei.jin@uq.edu.au if you encounter any issue when executing this program.
Thank you!