The Internet of Things(IoT)has aided in the development of new products and services.Due to the heterogeneity of IoT items and networks,traditional techniques cannot identify network risks.Rule-based solutions make it...The Internet of Things(IoT)has aided in the development of new products and services.Due to the heterogeneity of IoT items and networks,traditional techniques cannot identify network risks.Rule-based solutions make it challenging to secure and manage IoT devices and services due to their diversity.While the use of artificial intelligence eliminates the need to define rules,the training and retraining processes require additional processing power.This study proposes a methodology for analyzing constrained devices in IoT environments.We examined the relationship between different sized samples from the Kitsune dataset to simulate the Mirai attack on IoT devices.The training and retraining stages for the Mirai attack were also evaluated for accuracy.Various approaches are evaluated in smaller sample sizes to minimize training time on low-resource devices.Cross-validation was used to avoid overfitting classification methods during the learning process.We used the Bootstrapping technique to generate 1000,10000,and 100000 samples to examine the performance metrics of different-sized variations of the dataset.In this study,we demonstrated that a sample size of 10000 is sufficient for 99,56%accuracy and learning in the detection of Mirai attacks in IoT devices.展开更多
为挖掘需求侧资源响应潜力,文中提出一种计及多重需求响应的综合能源系统(integrated energy system,IES)多时间尺度低碳调度策略。首先,考虑到需求侧资源在不同时间尺度下的响应差异性,建立计及价格型和激励型的多重综合需求响应(integ...为挖掘需求侧资源响应潜力,文中提出一种计及多重需求响应的综合能源系统(integrated energy system,IES)多时间尺度低碳调度策略。首先,考虑到需求侧资源在不同时间尺度下的响应差异性,建立计及价格型和激励型的多重综合需求响应(integrated demand response,IDR)模型。然后,为减少源、荷预测误差对IES运行的影响,分别构建日前低碳经济调度模型和日内双时间尺度滚动优化平抑模型。最后,算例仿真设置不同场景进行对比分析。结果表明,相比传统IDR,多重IDR能有效挖掘用户响应潜力,提升系统经济性。此外,计及多重IDR的多时间尺度调度策略能有效缓解源、荷误差带来的功率波动并降低系统碳排放量,实现IES低碳、经济和稳定运行。展开更多
文摘The Internet of Things(IoT)has aided in the development of new products and services.Due to the heterogeneity of IoT items and networks,traditional techniques cannot identify network risks.Rule-based solutions make it challenging to secure and manage IoT devices and services due to their diversity.While the use of artificial intelligence eliminates the need to define rules,the training and retraining processes require additional processing power.This study proposes a methodology for analyzing constrained devices in IoT environments.We examined the relationship between different sized samples from the Kitsune dataset to simulate the Mirai attack on IoT devices.The training and retraining stages for the Mirai attack were also evaluated for accuracy.Various approaches are evaluated in smaller sample sizes to minimize training time on low-resource devices.Cross-validation was used to avoid overfitting classification methods during the learning process.We used the Bootstrapping technique to generate 1000,10000,and 100000 samples to examine the performance metrics of different-sized variations of the dataset.In this study,we demonstrated that a sample size of 10000 is sufficient for 99,56%accuracy and learning in the detection of Mirai attacks in IoT devices.
文摘为挖掘需求侧资源响应潜力,文中提出一种计及多重需求响应的综合能源系统(integrated energy system,IES)多时间尺度低碳调度策略。首先,考虑到需求侧资源在不同时间尺度下的响应差异性,建立计及价格型和激励型的多重综合需求响应(integrated demand response,IDR)模型。然后,为减少源、荷预测误差对IES运行的影响,分别构建日前低碳经济调度模型和日内双时间尺度滚动优化平抑模型。最后,算例仿真设置不同场景进行对比分析。结果表明,相比传统IDR,多重IDR能有效挖掘用户响应潜力,提升系统经济性。此外,计及多重IDR的多时间尺度调度策略能有效缓解源、荷误差带来的功率波动并降低系统碳排放量,实现IES低碳、经济和稳定运行。
文摘由于低照度图像具有对比度低、细节丢失严重、噪声大等缺点,现有的目标检测算法对低照度图像的检测效果不理想.为此,本文提出一种结合空间感知注意力机制和多尺度特征融合(Spatial-aware Attention Mechanism and Multi-Scale Feature Fusion,SAM-MSFF)的低照度目标检测方法 .该方法首先通过多尺度交互内存金字塔融合多尺度特征,增强低照度图像特征中的有效信息,并设置内存向量存储样本的特征,捕获样本之间的潜在关联性;然后,引入空间感知注意力机制获取特征在空间域的长距离上下文信息和局部信息,从而增强低照度图像中的目标特征,抑制背景信息和噪声的干扰;最后,利用多感受野增强模块扩张特征的感受野,对具有不同感受野的特征进行分组重加权计算,使检测网络根据输入的多尺度信息自适应地调整感受野的大小.在ExDark数据集上进行实验,本文方法的平均精度(mean Average Precision,mAP)达到77.04%,比现有的主流目标检测方法提高2.6%~14.34%.