Dr G.Sharmila Sujatha

https://doi.org/10.64823/ijter.2505005

The exponential growth of digital information exchange demands secure, efficient, and robust data encoding methods. This paper presents a unified data encoding framework integrating AES-GCM authenticated encryption with PBKDF2-based key derivation, Compress AI-driven learned compression, and a scalable multi- part encapsulation format with integrity verification via SHA-256 digests. The framework supports large payloads by dividing encrypted data into verifiable segments, enabling resilient storage and transmission. Optional image-quality enhancement using Real-ESRGAN or OpenCV EDSR is provided when the payload is visual media. Prototype evaluations show strong compression gains from Compress AI over classical JPEG+zlib baselines, strict integrity enforcement through AES-GCM tags, and accurate end-to-end reconstruction provided that all segments are available. The proposed approach is suited for privacy-preserving storage and controlled sharing in modern digital ecosystems.

https://doi.org/10.64823/ijter.2503019

Online payment fraud has been become a significant concern in financial sector, posing challenges for real-time detection and mitigation. This study gives us a machine learning-based fraud detection system designed for identifying fraudulent transactions both before and after their execution. A large transactional dataset is processed and filtered to focus on high-risk transaction types. A Random Forest classifier is implemented for fraud detection due to its robustness and high accuracy in handling imbalanced financial data using standard evaluation metrics. The proposed approach gives high accuracy, precision, and recall, particularly with ensemble models, indicating its effectiveness in enhancing fraud detection systems. The research contributes a deployed, user-interactive solution in Streamlit web interface.