Scholarlink

<p>Additive manufacturing (AM) combined with the innovation of the Internet of Things (IoT) has made manufacturing faster and more convenient than ever. However, increased connectivity has exposed systems to potential security breaches. To address this vulnerability, this study presents a novel methodology to monitor AM systems using a side-channel monitoring framework based on a power meter. The power meter measures the energy consumption of every G/M-code command. Concurrently, a machine learning (ML) model is trained to predict energy consumption for each G/M-code command in the absence of cyberattacks. Consequently, when attacks occur in the system, the energy consumption prediction errors increase, indicating the presence of malicious activity. Our method is validated with three different print models to ensure consistency and transferability. For all models, the prediction error increased by a factor of seven during attacks that altered the operating temperatures by 1 or 2 degrees. The results demonstrate the effectiveness of the proposed method in detecting attacks and anomalies in a timely manner.</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

Monitoring energy consumption for cyberattack detection in additive manufacturing systems

  • 期刊论文

  • 发表日期：

    2025年9月26日

查看全文
  • Xuan-Phat Truong,
  • Haemin Jeon,
  • Yi Wang,
  • Seong Hyeon Hong

摘要

Additive manufacturing (AM) combined with the innovation of the Internet of Things (IoT) has made manufacturing faster and more convenient than ever. However, increased connectivity has exposed systems to potential security breaches. To address this vulnerability, this study presents a novel methodology to monitor AM systems using a side-channel monitoring framework based on a power meter. The power meter measures the energy consumption of every G/M-code command. Concurrently, a machine learning (ML) model is trained to predict energy consumption for each G/M-code command in the absence of cyberattacks. Consequently, when attacks occur in the system, the energy consumption prediction errors increase, indicating the presence of malicious activity. Our method is validated with three different print models to ensure consistency and transferability. For all models, the prediction error increased by a factor of seven during attacks that altered the operating temperatures by 1 or 2 degrees. The results demonstrate the effectiveness of the proposed method in detecting attacks and anomalies in a timely manner.