Open Access

ARTICLE

A Blockchain-Based Adaptive Security Framework with Real-Time Incident Response and Usability Feedback for Non-Expert Users

Mosammat Jannatul Kobra¹, Muhammad Rashid Majeed^2,*, Md Owahedur Rahman¹

1 School of Electronics and Information Engineering, Nanjing University of Information Science & Technology, Nanjing, China
2 School of Computer Science, Nanjing University of Information Science & Technology, Nanjing, China

* Corresponding Author: Muhammad Rashid Majeed. Email:

Journal of Blockchain and Intelligent Computing 2026, 2, 27-44. https://doi.org/10.32604/jbic.2026.081492

Received 04 March 2026; Accepted 20 April 2026; Issue published 13 May 2026

Abstract

The proposed study introduces a blockchain-based framework for an adaptive security solution with real-time incident response and usability feedback for non-expert users. Traditional security solutions are often designed with static, opaque policies, which makes them complex. These issues make them less effective in dealing with complex environments. Thus, to make them more effective, the proposed framework introduces supervised machine learning for attack classification, unsupervised machine learning for anomaly detection, a risk-aware, adaptive policy engine, and a lightweight, tamper-evident, hash-linked ledger for auditable decision-making. The proposed framework uses a Random Forest classifier for BENIGN/ATTACK classification, and an Isolation Forest module for anomaly detection. The proposed framework also uses an adaptive policy engine to decide whether to use HIGH or LOW security mode based on a combined risk score calculated from the probability of an attack and the rate of anomalies. Moreover, the proposed framework incorporates usability feedback through interaction steps, user errors, and setup time, making it suitable for non-expert users. The proposed framework was implemented using the CICIDS2017 benchmark dataset for intrusion detection, which contains 2,830,743 flows with 70 numerical features from 8 CSV files. The proposed framework achieved an accuracy of 0.9789, precision of 0.9753, recall of 0.9793, F1-score of 0.9773, and ROC-AUC of 0.9799 using the Random Forest classifier. The proposed framework achieved an anomaly rate of 0.1456 using the Isolation Forest module and a risk score of 0.1824 using the adaptive policy engine, which decides whether to use LOW or HIGH security mode. Moreover, the proposed framework enabled a simple interaction process, making it suitable for non-expert users.

---

Keywords

---