A Hybrid CNN-RSA Framework for Secure PDF Document Classification and Storage in Distributed Cloud Environments

Abstract

The increasing adoption of cloud computing has improved the storage, accessibility, and management of digital documents. However, the security of Portable Document Format (PDF) files remains a major concern because of unauthorized access, data leakage, malicious document attacks, and privacy violations. Traditional cloud security mechanisms often focus either on encryption or on document classification, leaving a gap in providing an integrated solution that supports both intelligent document analysis and secure storage. This paper presents a Hybrid PDF Security Trainer Framework that integrates Term Frequency–Inverse Document Frequency (TF-IDF), Convolutional Neural Networks (CNN), and Rivest–Shamir–Adleman (RSA) encryption to provide a comprehensive security mechanism for PDF documents stored in distributed cloud environments. The proposed framework first extracts text from uploaded PDF documents and then converts the text into numerical feature vectors using TF-IDF. The generated features are processed by a CNN model that classifies documents according to their security characteristics and detects potentially sensitive or malicious content. Based on the classification outcome, an appropriate security policy is selected and RSA encryption is applied before the document is stored in a distributed cloud infrastructure that supports secure access and controlled retrieval. The framework combines intelligent machine learning with cryptographic protection to improve document confidentiality while maintaining efficient storage and retrieval. The integrated architecture enhances confidentiality, reduces unauthorized access, and provides a practical solution for secure document management in modern cloud computing environments.