RE-DACT: Adaptive Redaction and Anonymization Tool Using Machine-Learning

“RE-DACT” is a safe and user-friendly redaction tool for customizable redaction, masking, and anonymization using a user-defined gradational scale. With the help of NLP and ML, the tool gives users the opportunity to specify data elements that are going to be redacted, ranging from simple name removal and more complex anonymization techniques to the generation of fully synthetic data, all while keeping the structure of the content unchanged. This tool is usable both online and offline in its web- based interface and takes into account the common input formats for text files, images, as well as PDFs. Over time, RE-DACT learns to make realistic synthetic datasets suitable for training, testing, as well as commercial applications without compromising privacy. It therefore holds robust data security with the minimum retention of data and prevents unauthorized access to sensitive information. The secure coding and scalable solutions support real-world applications. Performance is measured in terms of precision, recall, F1 score, redaction efficacy, speed, and ease of use. “The tool incorporates anonymization techniques, such as k-anonymity and synthetic data generation, to ensure privacy. It makes use of advanced encryption methods and ensures safe placeholders to guard sensitive information from unauthorized access.”