🔍

←

Our strategy of data redaction involved custom-built solutions that addressed their several use cases. However the basic process outlines as below:

Data Collection: The first step is to collect the relevant data that needs to be redacted. This may include documents, emails, databases, or any other sources containing sensitive information.

Text Extraction: Once the data is collected, the next step is to extract the text from the different sources. This can involve converting documents into machine-readable formats or extracting text from structured databases.

Processing: The extracted text is then processed using various techniques to identify and redact sensitive information. Custom Regular Expression (REGEX) patterns can be used to search for specific patterns or formats of sensitive data like social security numbers or credit card numbers. Named Entity Recognition (NER) algorithms can identify entities like names, addresses, or medical terms. Rule-based logic and deny lists can be applied to identify and redact specific keywords or phrases.

Algorithm (BERT, Fine-tune BERT): Advanced algorithms like Bidirectional Encoder Representations from Transformers (BERT) can be utilized for more complex redaction tasks. BERT-based models can be fine-tuned on specific datasets to improve the accuracy and effectiveness of redaction, especially in cases where context and semantics play a crucial role.

Output: The final step is to generate the redacted output. This can be in the form of updated documents, databases, or reports where the sensitive information has been replaced or masked. The output should ensure that the redacted data is effectively removed or obfuscated, while maintaining the integrity and structure of the original information.