Hide 'n Seek: Personal Identifiable Info (PII) Redactor

With the help of Natural Language Processing AI Model, we would identify any PII details from PDF documents and redact them. This also includes a tipping feature using XRP test network.

Comment

Inspiration

With governments across the globe passing stricter Data Protection and Privacy Preservation laws, Institutions and companies across the globe face significant administrative overhead to comply with them. Therefore we want to build Ai powered cost effective Privacy Preservation solutions that can assist individuals in complying with these laws. They have to be democratized, inexpensive, simple and accessible for wider adaptation.

What it does

With the help of Natural Language Processing AI Model (Spacy), we would identify any PII from PDF documents and redact them. It also included a tipping feature using XRP testnet.

How we built it

We had first started by brainstorming ideas in the privacy preservation space. Once we had a skeleton plan ready, we made a choice on focusing on the core functionality of the problem. We prioritized developing a fully functional backend engine (our MVP) that proves viability of our service. We used HyperCycle AWS Ec2 instance to deploy our ML model (Spacy) as well as create a PDF parser that reads and tokenize it. Using these tokens, the model determines if its PII or not. Parallel processing techniques were used to reduce latency of it. We have also used XRPL python library to introduce a brand new tipping mechanism within the service that gets settled once a transaction is concludes. This was built using python and FastAPI service.

Challenges we ran into

  • FastAPI and XRPL transactions had async issues that broke our pipeline
  • The PDF document wasn't rendering to image when tested on FastAPI
  • Understanding the logic for Processing Units and come up with a new measure for our service

Accomplishments that we're proud of

  • Being able to learn and use XRPL framework for the very first time
  • Work and deploy AI model on HyperCycle AWS Ec2 instance
  • Having a fully functional backend service that can be deployed immediately

What we learned

Being the first crypto hackathon, we have learned:

  • How to deploy AI model on EC2 instance
  • How to write smart contracts using XRPL libraries

What's next for Hide 'n Seek: Personal Identifiable Information (PII) Redactor Service

  • Make the service run on the ARM-based edge-computing hardware of the HyperAiBox
  • Build a simple UI for consumers to use this service
  • Make money out of this service

Built With

Share this project:

Updates