In network-connected UAV(NCUAV) communication systems, user authentication is replaced by platform identity authentication and integrity check because many NC-UAVs are operated without human intervention. Direct anony...In network-connected UAV(NCUAV) communication systems, user authentication is replaced by platform identity authentication and integrity check because many NC-UAVs are operated without human intervention. Direct anonymous attestation(DAA) is an attractive cryptographic scheme that provides an elegant balance between platform authentication and anonymity. However, because of the low-level computing capability and limited transmission bandwidth in UAV, the existing DAA schemes are not suitable for NC-UAV communication systems. In this paper, we propose an enhanced DAA scheme with mutual authentication(MA-DAA scheme), which meets the security requirements of NC-UAV communication systems. The proposed MA-DAA scheme, which is based on asymmetric pairings, bundles the identities of trusted platform module(TPM) and Host to solve the malicious module changing attacks. Credential randomization, batch proof and verification, and mutual authentication are realized in the MA-DAA scheme. The computational workload in TPM and Host is reduced in order to meet the low computation and resource requirements in TPM and Host.The entire scheme and protocols are presented,and the security and efficiency of the proposed MA-DAA scheme are proved and analyzed.Our experiment results also confirm the high efficiency of the proposed scheme.展开更多
杂草作为一种常见的农业问题,对农作物的生长造成比较严重的影响,控制和管理杂草是农业生产活动中的重要一环。近年来,随着无人机技术和人工智能技术的快速发展,基于无人机平台的特定区域杂草管理是目前除草作业的主流研究,而精确高效...杂草作为一种常见的农业问题,对农作物的生长造成比较严重的影响,控制和管理杂草是农业生产活动中的重要一环。近年来,随着无人机技术和人工智能技术的快速发展,基于无人机平台的特定区域杂草管理是目前除草作业的主流研究,而精确高效地对田间杂草进行识别和检测是实现自动化杂草管理的重要前提。但高效的识别模型往往意味着大量的农业数据。为了降低对农业标签数据的依赖性,该研究提出了一种UANP-MT(uncertainty aware and network perturbed mean teacher)的半监督语义分割网络。该模型基于PSPNet结构与MT(mean teacher)的思想,首先通过对教师网络做扩增输出,令该部分做出若干次推理并取其均值,以此来保证网络预测的鲁棒性,其次在网络的一致性学习部分构建不确定性系数来约束不同网络间的输出差异,提高预测的置信度和可靠性,从而提高模型的识别准确度。为了验证所提出的模型的有效性,设计消融试验,包括对网络参数的取值设置,特征提取网络backbone的选取,以及在不同数据量的数据集下对模型进行性能测试,试验过程中确定了模型的一些最佳的参数设置。结果表明,在与原监督网络的对比试验中,在所提出的UANP-MT模型在标签数据低于原监督网络的前提下,其F1分数,像素精确度PA(pixel accuracy)以及交并比Iou(intersection over union)3个评估指标或皆比原监督网络更高,性能更优。此外,在与常用的语义分割模型的对比中,UANP-MT也体现出了其优越性,在1/4数据集的标签数据量参与训练的情况下F1分数为81.83%,像素准确度为95.84%,交并比为90.70%。评估指标分别优于次之的Deeplabv3+模型4.71,7.94,8.27个百分点。该模型能够较好地在低标签数据量情况下对杂草数据集做出高质量的检测和识别,极大地减少物力和时间成本,对后续开发无人机平台的自动化除草作业有一定的参考作用。展开更多
基金supported in part by the European Commission Marie Curie IRSES project "AdvIOT"the National Natural Science Foundation of China (NSFC) under grant No.61372103
文摘In network-connected UAV(NCUAV) communication systems, user authentication is replaced by platform identity authentication and integrity check because many NC-UAVs are operated without human intervention. Direct anonymous attestation(DAA) is an attractive cryptographic scheme that provides an elegant balance between platform authentication and anonymity. However, because of the low-level computing capability and limited transmission bandwidth in UAV, the existing DAA schemes are not suitable for NC-UAV communication systems. In this paper, we propose an enhanced DAA scheme with mutual authentication(MA-DAA scheme), which meets the security requirements of NC-UAV communication systems. The proposed MA-DAA scheme, which is based on asymmetric pairings, bundles the identities of trusted platform module(TPM) and Host to solve the malicious module changing attacks. Credential randomization, batch proof and verification, and mutual authentication are realized in the MA-DAA scheme. The computational workload in TPM and Host is reduced in order to meet the low computation and resource requirements in TPM and Host.The entire scheme and protocols are presented,and the security and efficiency of the proposed MA-DAA scheme are proved and analyzed.Our experiment results also confirm the high efficiency of the proposed scheme.
文摘杂草作为一种常见的农业问题,对农作物的生长造成比较严重的影响,控制和管理杂草是农业生产活动中的重要一环。近年来,随着无人机技术和人工智能技术的快速发展,基于无人机平台的特定区域杂草管理是目前除草作业的主流研究,而精确高效地对田间杂草进行识别和检测是实现自动化杂草管理的重要前提。但高效的识别模型往往意味着大量的农业数据。为了降低对农业标签数据的依赖性,该研究提出了一种UANP-MT(uncertainty aware and network perturbed mean teacher)的半监督语义分割网络。该模型基于PSPNet结构与MT(mean teacher)的思想,首先通过对教师网络做扩增输出,令该部分做出若干次推理并取其均值,以此来保证网络预测的鲁棒性,其次在网络的一致性学习部分构建不确定性系数来约束不同网络间的输出差异,提高预测的置信度和可靠性,从而提高模型的识别准确度。为了验证所提出的模型的有效性,设计消融试验,包括对网络参数的取值设置,特征提取网络backbone的选取,以及在不同数据量的数据集下对模型进行性能测试,试验过程中确定了模型的一些最佳的参数设置。结果表明,在与原监督网络的对比试验中,在所提出的UANP-MT模型在标签数据低于原监督网络的前提下,其F1分数,像素精确度PA(pixel accuracy)以及交并比Iou(intersection over union)3个评估指标或皆比原监督网络更高,性能更优。此外,在与常用的语义分割模型的对比中,UANP-MT也体现出了其优越性,在1/4数据集的标签数据量参与训练的情况下F1分数为81.83%,像素准确度为95.84%,交并比为90.70%。评估指标分别优于次之的Deeplabv3+模型4.71,7.94,8.27个百分点。该模型能够较好地在低标签数据量情况下对杂草数据集做出高质量的检测和识别,极大地减少物力和时间成本,对后续开发无人机平台的自动化除草作业有一定的参考作用。
