IOT-Based Malware Detection Framework Using Polymorphic AES and Blockchain with Proof of Work Mechanism

In recent years, the attack detection framework used by IoT systems for monitoring has been used to collect and analyze data related to detecting user behavior, predicting potential attacks, and responding in a predetermined manner. This manuscript introduces the innovative multi-agent system and blockchain (BC) technology to enhance data security and detection capabilities in an attack. The primary data sources are initially collected from the freely accessible dataset obtained from the Kaggle platform. The Multiple Imputation-Chained Equations preprocessing handles missing values so that the data's integrity is not compromised. After preprocessing, the data is encrypted using Polymorphic Advanced Encryption Standard (AES), safeguarding private information. Then, the encrypted data is safely stored in a BC environment using a Proof of Work (PoW) mechanism that guarantees validity and immutability in recorded data and resistance to unauthorized modification. Then, the proposed framework deploys the Multi-Scale Channel Attention Residual Network (MSC-Att-ResNet), which analyzes the encrypted data for malicious patterns. To detect intrusions, the Clouded Leopard Optimizer (CdLO) algorithm continually monitors the network activity with fixed predefined threshold values to identify potential attacks. The proposed method is simulated via the Python environment. The evaluation metrics accuracy, false negative rate (FNR), encryption and decryption time, and Matthew’s correlation rate (MCC) are evaluated through a comparative analysis with existing approaches. The overall accuracy of 99.01%, MCC of 97.16%, FNR of 2.60, ET, and DT of 174.79ms are obtained by the proposed framework.