Biblio

Redaction of personal, confidential and sensitive information from documents is becoming increasingly important for individuals and organizations. In past years, there have been many well-publicized cases of data leaks from various popular companies. When the data contains sensitive information, these leaks pose a serious threat. To protect and conceal sensitive information, many companies have policies and laws about processing and sanitizing sensitive information in business documents.The traditional approach of manually finding and matching millions of words and then redacting is slow and error-prone. This paper examines different models to automate the identification and redaction of personal and sensitive information contained within the documents using named entity recognition. Sensitive entities example person’s name, bank account details or Aadhaar numbers targeted for redaction, are recognized based on the file’s content, providing users with an interactive approach to redact the documents by changing selected sensitive terms.

Guaranteeing a certain level of user privacy in an arbitrary piece of text is a challenging issue. However, with this challenge comes the potential of unlocking access to vast data stores for training machine learning models and supporting data driven decisions. We address this problem through the lens of dx-privacy, a generalization of Differential Privacy to non Hamming distance metrics. In this work, we explore word representations in Hyperbolic space as a means of preserving privacy in text. We provide a proof satisfying dx-privacy, then we define a probability distribution in Hyperbolic space and describe a way to sample from it in high dimensions. Privacy is provided by perturbing vector representations of words in high dimensional Hyperbolic space to obtain a semantic generalization. We conduct a series of experiments to demonstrate the tradeoff between privacy and utility. Our privacy experiments illustrate protections against an authorship attribution algorithm while our utility experiments highlight the minimal impact of our perturbations on several downstream machine learning models. Compared to the Euclidean baseline, we observe \textbackslashtextgreater 20x greater guarantees on expected privacy against comparable worst case statistics.