1. Introduction

The protection granted to an automated information system, focused on preserving integrity, availability, and confidentiality of system resources, describes computer security succinctly. Technological advancements have significantly contributed to the easy and speedy transfer of data. Data constitutes a key factor for conducting business activities, hence, the necessity to share data among multiple stakeholders such as human resource individuals, business colleagues, and other clients.

Sensitive data leakage or loss has emerged as the major threat that organizations grapple with today. With nearly all business operations depending on extensive sharing of sensitive data, a data leakage event occurring inadvertently or maliciously compromises an organization's reputation and customers' trust. Strategies to minimize this risk usually incorporate Data Leakage Prevention (DLP) solutions as a protection mechanism. This technology supersedes traditional protection technologies like firewalls and IDS protocols. DLP solutions are used to protect all types of data, be it data in use, at rest or in transition.

This blog presents a detailed explanation about DLP approach, how it distinguishes between normal and suspicious activities, and a proposed data training model for added security.

2. Overview of DLP Approach

DLP model characteristically distinguishes routine activity from suspicious activity, and initiates either detection, i.e., raises alert if any doubtful activity is detected or prevention, i.e., blocks undesirable activities. The approach to model construction ideally works best for the proposed structure. The proposed data fitting model framework offers numerous benefits compared to existing solutions for DLP. It tailors itself to the user's behavior, detects unknown and insider attacks, provides data misuse control, and allows simultaneous data access. Its integration with user identity allows the organization to implement data protection policy based on the identity and the role of the user.

3. Data Training Model

The proposed DLP model achieves the objective of data security by employing a machine learning approach. The model closely analyzes each user access to the organization's data and flags any trends out of the ordinary activity, enabling the organization to take suitable actions like imposing access restriction on sensitive data for enhanced security.

The proposed model uses a simple piecewise linear function for learning and training the model. The dataset of user activity from 2014 to 2018 is used to train the model and predict the trend of a particular user after 2018. The model calculates the error existing between the actual and predicted value to determine its own accuracy. If the error exceeds a defined upper limit, the model raises an alert to prevent or restrict the user's access to an organization's data.

4. Conclusion

The proposed DLP approach offers an efficient and timely solution to prevent data leakage in organizations. It ensures that users without appropriate permission cannot access sensitive data and provides protection to sensitive data which might be shared accidentally. Moreover, by making effective use of statistical analysis, the approach is able to forecast data access possibilities of any user in the future based on the access to data in the past. The utilization of data training models is an added advantage in enhancing the precision of the approach. Therefore, implementing robust DLP solutions is crucial for organizations in preventing data leaks and protecting their data resources.