k-scale: k-Anonymizing Millions of Trajectories

This repository contains the implementation, evaluation pipeline, and benchmarking notebooks for the paper:

“k-scale: k-Anonymizing Millions of Trajectories”
by Abhishek Kumar Mishra and Marco Fiore, published at the IEEE INFOCOM 2026.

The work introduces k-scale, a scalable framework for efficiently achieving trajectory k-anonymity on datasets containing millions of mobility traces while preserving strong analytical utility.

---

Paper

• HAL Archive: https://hal.science/hal-05407897
• PDF: https://hal.science/hal-05407897v1/file/Efficiently_k_anonymizing_a_Million_Trajectories___IEEE_INFOCOM__HAL_Version_.pdf

If you use this repository, please cite the paper below.

---

Repository Structure

.
├── Benchmarks/
├── DatasetGIS/
├── Evaluation/
├── Results/
├── Source/
└── README.md

---

Folder Description

Source/

Contains the core implementation of the k-scale framework.

---

Benchmarks/

Contains notebooks and scripts used to compare k-scale against state-of-the-art trajectory anonymization baselines.

---

DatasetGIS/

Contains example notebooks demonstrating how to preprocess raw mobility datasets (e.g., CDR datasets).

---

Evaluation/

Contains the notebooks used to reproduce the experimental results reported in the paper.

---

Results/

Contains the generated experimental outputs.

---

Reproducing the Paper Results

The general workflow is:

Raw Dataset
    ↓
DatasetGIS preprocessing
    ↓
k-scale anonymization (Source/)
    ↓
Evaluation notebooks
    ↓
Results generation

---

Requirements

The implementation is primarily based on Python notebooks and scientific Python libraries.

Typical dependencies include:

numpy
pandas
scikit-learn
geopandas
matplotlib
scipy
jupyter

Depending on the evaluation notebook, additional packages may be required.

---

Citation

Please cite the following paper if you use this repository or build upon this work:

@inproceedings{mishra:hal-05407897,
  title     = {k-scale: k-Anonymizing Millions of Trajectories},
  author    = {Mishra, Abhishek Kumar and Fiore, Marco},
  booktitle = {IEEE INFOCOM 2026 Conference Proceedings},
  address   = {Tokyo, Japan},
  year      = {2026},
  month     = may,
  url       = {https://hal.science/hal-05407897},
  pdf       = {https://hal.science/hal-05407897v1/file/Efficiently_k_anonymizing_a_Million_Trajectories___IEEE_INFOCOM__HAL_Version_.pdf},
  hal_id    = {hal-05407897},
  hal_version = {v1}
}

---

Acknowledgment

This repository accompanies the research work presented at IEEE INFOCOM 2026 and is intended to support reproducible research in large-scale mobility privacy and trajectory anonymization.