A Proposal for Data Breach Detection in Organizations Based on User Behavior

Abstract

Data breach has become a big problem for organizations, as the consequences can range from loss of reputation to financial loss. A data breach occurs through outsiders and insiders; however, threats from insiders are the most common and, at the same time, the most difficult to prevent. Data loss detection systems are increasingly implemented in organizations to protect information with techniques like content-based and context-based checking. Machine learning techniques have proven to be useful for data breach detection. In this work, a statistical analysis of data breach incidents is presented. Also, a user behavior characterization is made, mainly based on incidents reported by various organizations. Part of this characterization is used to create a machine learning model with a long short-term memory network with an autoencoder, in order to identify anomalies in user behavior to detect data breaches from insiders.