An Anomalous Behavior Detection Method Based on Power Analysis Utilizing Steady State Power Waveform Predicted by LSTM
An Anomalous Behavior Detection Method Based on Power Analysis Utilizing Steady State Power Waveform Predicted by LSTM
复制标题
一种利用 LSTM 预测的稳态功率波形进行功率分析的异常行为检测方法
DOI：
10.1109/iolts52814.2021.9486706
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Togawa Nozomu
中科院分区：
文献类型：
--
作者：
Takasaki Kazunari;Kida Ryoichi;Togawa Nozomu
近年来，随着物联网（IoT）设备的迅速普及，硬件安全问题也随之出现。功率分析是检测异常操作的方法之一，但很难将其应用于运行操作系统和各种软件程序的物联网设备，因此其功率波形变得更加复杂。在本文中，我们提出了一种异常行为检测方法，该方法利用由LSTM（长短期记忆）生成的稳态功率波形提取的特定于应用的功率行为。所提出的方法是基于通过预测稳态功率波形来提取特定于应用的功率行为。此时，通过使用LSTM，我们可以有效地预测稳态功率波形，即使它们包括一个或多个循环波形和/或由许多复杂波形组成。在单板机上实现了3个正常应用程序和1个异常应用程序，实验结果表明，该方法能够成功地检测出异常应用程序的异常功耗行为，而现有方法不能。
Hardware security issues have emerged in recent years as Internet of Things (IoT) devices have rapidly spread. Power analysis is one of the methods to detect anomalous operations, but it is hard to apply it to IoT devices where an operating system and various software programs are running and hence its power waveforms become more complex. In this paper, we propose an anomalous behavior detection method utilizing application-specific power behaviors extracted by steady-state power waveform, which is generated by LSTM (long short-term memory). The proposed method is based on extracting application-specific power behaviors by predicting steady-state power waveforms. At that time, by using LSTM, we can effectively predict steady-state power waveforms, even if they include one or more cycled waveforms and/or they are composed of many complex waveforms. In the experiment, we implement three normal application programs and one anomalous application program on a single board computer and apply the proposed method to it. The experimental results demonstrate that the proposed method successfully detects the anomalous power behavior of an anomalous application program, while the existing method cannot.