Adaptive cryptography: a transformer neural network-based approach for anomaly detection and secure messaging with signReencryption
The increasing sophistication of cyber threats, particularly in decentralized and resource-constrained environments such as the Internet of Things (IoT), demands adaptive and efficient security solutions. This study introduces SignReencryption, a unified framework that integrates signcryption, proxy...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Year Project / Dissertation / Thesis |
| Published: |
2025
|
| Subjects: | |
| Online Access: | http://eprints.utar.edu.my/7281/1/SE_2105387_FYP_Report_%2D_TeeJunnJeh_TEE_JUNN_JEH.pdf http://eprints.utar.edu.my/7281/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The increasing sophistication of cyber threats, particularly in decentralized and resource-constrained environments such as the Internet of Things (IoT), demands adaptive and efficient security solutions. This study introduces SignReencryption, a unified framework that integrates signcryption, proxy re-encryption (PRE), and Transformer-based intrusion detection to deliver both cryptographic assurance and intelligent adaptability. Signcryption ensures confidentiality and authenticity in a single lightweight operation, while PRE enables scalable, fine-grained access control without exposing plaintext. A TabTransformer-based intrusion detection system complements these cryptographic mechanisms, achieving classification accuracies of 94% on CICIDS2017, 99% on CIDDS-001, and 97% on NSL-KDD, with particular strength in detecting minority attack classes traditionally overlooked by baseline models. Optuna-driven hyperparameter optimization revealed dataset-specific configurations, demonstrating the adaptability of the TabTransformer across heterogeneous traffic distributions. Experimental evaluation further shows that SignReencryption reduces ciphertext expansion by up to 50% and lowers per-message execution time by nearly half compared to conventional Sign-Then-Encrypt schemes, confirming its practicality for real-time and bandwidth-limited environments such as intelligent transportation systems. Overall, the framework advances intrusion detection by uniting cryptographic efficiency with adaptive intelligence, offering a scalable, resilient, and operationally viable defense model for modern cybersecurity challenges.
Keywords: Signcryption; Cryptography; Transformer Neural Network; Intrusion Detection System; Internet of Things
Subject Area: QA75.5-76.95 |
|---|