Hiding Patterns With Gaps in Sequential Data

ABSTRACT String sanitization addresses the challenge of removing sensitive patterns from text data while retaining the usefulness of the remaining content. The task becomes especially demanding when sensitive patterns may span variable length gaps—a scenario common in fields like bioinformatics, web analysis, and network traffic monitoring. In this work, we formalize the Pattern Hide with Gaps (PHG) problem, extending traditional string sanitization to handle VLG patterns. To solve PHG, we introduce three novel sanitization algorithms that balance different aspects of data utility: the first algorithm rapidly removes all sensitive patterns to achieve basic sanitization; the second carefully selects replacements to minimize overall distortion; and the third algorithm focuses on reducing the loss of frequent patterns to enhance the accuracy of subsequent frequent pattern mining tasks. Extensive experiments demonstrate that our methods run efficiently and successfully maintain high data utility.