REDACT
REDACT is an info-sec tool that automates redaction with minimal user interaction. It utilizes spaCy NLP alongside BeRT model through TensorFlow and Hugging face Transformers.
Created on 7th February 2025
•
REDACT
REDACT is an info-sec tool that automates redaction with minimal user interaction. It utilizes spaCy NLP alongside BeRT model through TensorFlow and Hugging face Transformers.
The problem REDACT solves
REDACT (Rapid Efficient Data Anonymozation & Content Transformation) is an smart redaction tool designed to efficiently redact sensitive data from docs, images, audio and various other file formats for minimizing manual effort.
Video Demo | Examples | Test Material | Source Code
Challenges I ran into
Multilingual support with mBeRT. Now enables tokenization in ~100 languages (previously data was recognized and translated to english before processing. Translation methodology excluded bcz loss of context occoured).
Stress-testing model(s) with CUDA for faster model inference and offloading CPU work. Multiprocessing pool monitored with perf & strace.
Discussion
Builders also viewed
See more projects on Devfolio