Hack Five
Hack Five

Open Source AI Hackathon 2024

GenAI makers and creators contest and showcase

Tickets

Loading…

All submissions

Rumanu

@rumanu

Applying Differential Privacy to India's Census Data

Submitted Feb 2, 2024

Category: AI for India

Bringing Technically Ensured Privacy to Indian Data in the Public Domain

Modelled after The US Government’s attempt to release 2020 Census data in a differentially private manner, the project aims to successively apply Differential Privacy to 0.001%, 0.01% and 0.1% of India’s census data to be able to release such personal data that allows us to obtain aggregate statistics about the trends in the population without compromising entirely on individual privacy for the members involved. The attemp wants to define mathematically rigourous limits for the level of guaranteed privacy in such instances so make the approach measurable instead of qualitative in nature.

This is an experimental approach to estimate the feasibility of attempting this at scale and checking to see if insights drawn still serve the purpose of drawing demographic insights that allow the governance to gauge and serve the needs of its population in all its diversity.

The project also aims to assess whether the analysis thus drawn would be compliant with the terms proposed by the Personal Data Protection Bill of India, if such a dataset were acquired by consumers of a startup as a starting point to see the efficacy of DP as a solution to business analysis needs.

The solution will involve using PyTorch’s Opacus, Tensorflow’s Privacy tools as well as OpenMined’s PyDP and Harvard’s OpenDP libraries, all open source, as the solution demands, based on the merits and demerits of each library’s speed, design and ease of use.

All submissions

Comments

Login to leave a comment

  • A

    Akshobhya

    @akshobhya_j Editor & Promoter

    Hello @rumanu,

    The hack day for The Fifth Elephant Open Source AI Hackathon officially kicked off on Saturday, 3rd February! If you haven't started already, now is the perfect moment to dive in and begin building your project.

    Your project needs additional information to be shortlisted in The Fifth Elephant Open Source AI Hackathon.

    Don't miss this opportunity to collaborate, create, and possibly win one of the five prizes of INR. 1,00,000 each. Submit your idea, join the conversation with our mentors, and make the most of this unique hackathon experience.

    All the best and happy hacking!

    Posted 1 year ago
  • A

    Akshobhya

    @akshobhya_j Editor & Promoter

    Critical Assessment:

    1. The step-wise approach to apply Differential Privacy at different scales allows for a comprehensive assessment.
    2. The emphasis on mathematically rigorous privacy limits adds robustness to the privacy preservation methodology.
    3. Evaluating compliance with the Personal Data Protection Bill of India demonstrates awareness of legal considerations.
    4. The choice of diverse open-source libraries reflects adaptability and consideration for specific requirements.

    Considerations:

    1. The experimental nature of this project requires careful evaluation of the potential impact on the accuracy and utility of the census data at different privacy levels.
    2. Ensuring the scalability of applying Differential Privacy to the entire Indian census dataset is a critical technical challenge.
    3. Integration of multiple open-source libraries may introduce complexities in the implementation and maintenance of the solution.
    4. Ethical considerations in handling potentially sensitive demographic information should be a priority throughout the project.

    Overall Impression:

    The project demonstrates a thorough understanding of Differential Privacy and its potential application to Indian census data. The integration of multiple open-source libraries reflects a pragmatic approach. However, technical and ethical considerations should be carefully addressed to ensure the viability and ethical soundness of the implemented solution.

    Posted 1 year ago
Hybrid access (members only)

Hosted by

Hack Five
Hack Five
The Fifth Elephant hackathons

Supported by

Host

The Fifth Elephant
The Fifth Elephant
Jump starting better data engineering and AI futures
Meta
Meta

Venue host

H
Hasura
Welcome to the events page for events hosted at The Terrace @ Hasura. more

Partner

Microsoft for Startup's
Microsoft for Startup's
Providing all founders, at any stage, with free resources to build a successful startup.