Scholarlink

<p>Malware detection is the process which identifies, blocks and prevents the harmful effects of malware by using a set of defensive techniques and technologies. Federated Learning (FL) is a sub-field of Machine Learning (ML), wherein training data remains decentralized. Here, the global model learns by utilizing the local update aggregation. However, training data can be manipulated or corrupted by malicious activities, which may affect the locally computed updates. Therefore, detecting malevolent IoT devices is crucial to prevent such issues. Here, Federated learning-based Deep Belief Network Fused SpinalNet (FedL_DBNFSpinalNet) is presented for Malware Detection (MD) in IoT devices using FL. This framework comprises global and local modes, while the designed model contains clients and servers. In the local model, acquisition of data, pre-processing of data, model training and model evaluation are conducted. The pre-processing is done employing Z-score Normalization (ZN) and model training is performed to detect malware using DBNFSpinalNet. Deep Belief Network (DBN) and SpinalNet are combined to form DBNFSpinalNet. In the global model, model aggregation and model initialization are accomplished. The model aggregation is carried out using the Conditional Autoregressive Value at Risk (CAViaR) model. Additionally, FedL_DBNFSpinalNet obtained high accuracy, mean average precision, recall, and F1-score values of about 88.7%, 87.2%, 91.3%, and 91%, as well as minimum False Positive Rate (FPR), loss, Mean Squared Error (MSE), and Root Mean Squared Error (RMSE), values about 0.087, 0.049, 0.068 and 0.026.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

FedL_DBNFSpinalNet based malware detection in IoT devices

  • 期刊论文

  • 发表日期：

    2025年2月8日

查看全文
  • R. Bhavani,
  • Veeramalai Sankaradass

摘要

Malware detection is the process which identifies, blocks and prevents the harmful effects of malware by using a set of defensive techniques and technologies. Federated Learning (FL) is a sub-field of Machine Learning (ML), wherein training data remains decentralized. Here, the global model learns by utilizing the local update aggregation. However, training data can be manipulated or corrupted by malicious activities, which may affect the locally computed updates. Therefore, detecting malevolent IoT devices is crucial to prevent such issues. Here, Federated learning-based Deep Belief Network Fused SpinalNet (FedL_DBNFSpinalNet) is presented for Malware Detection (MD) in IoT devices using FL. This framework comprises global and local modes, while the designed model contains clients and servers. In the local model, acquisition of data, pre-processing of data, model training and model evaluation are conducted. The pre-processing is done employing Z-score Normalization (ZN) and model training is performed to detect malware using DBNFSpinalNet. Deep Belief Network (DBN) and SpinalNet are combined to form DBNFSpinalNet. In the global model, model aggregation and model initialization are accomplished. The model aggregation is carried out using the Conditional Autoregressive Value at Risk (CAViaR) model. Additionally, FedL_DBNFSpinalNet obtained high accuracy, mean average precision, recall, and F1-score values of about 88.7%, 87.2%, 91.3%, and 91%, as well as minimum False Positive Rate (FPR), loss, Mean Squared Error (MSE), and Root Mean Squared Error (RMSE), values about 0.087, 0.049, 0.068 and 0.026.