本文设计了一种新的深度神经网络(New Deep Neural Network,NDNN)模型,并将其应用到入侵检测系统中.NDNN以其突出的特征学习能力充分学习训练数据的特征,在输出层,NDNN通过Softmax分类器对网络攻击报文与正常报文数据进行识别和分类,检...本文设计了一种新的深度神经网络(New Deep Neural Network,NDNN)模型,并将其应用到入侵检测系统中.NDNN以其突出的特征学习能力充分学习训练数据的特征,在输出层,NDNN通过Softmax分类器对网络攻击报文与正常报文数据进行识别和分类,检测异常报文与入侵攻击.实验通过对KDD Cup 99数据集进行仿真,实验结果表明本文设计的基于NDNN的入侵检测系统模型,进一步提高了入侵检测系统的精度,增强了网络的安全性.展开更多
In order to improve the network performance furthermore, a routing algorithm for 2D-Torus is investigated from the standpoint of load balance for virtual channels. The 2D-Torus network is divided into two virtual netw...In order to improve the network performance furthermore, a routing algorithm for 2D-Torus is investigated from the standpoint of load balance for virtual channels. The 2D-Torus network is divided into two virtual networks and each physical channel is split into three virtual channels. A novel virtual channel allocation policy and a routing algorithm are proposed, in which traffic load is distributed to those three virtual channels in a more load-balanced manner by introducing a random parameter. Simulations of the proposed algorithm are developed with a SystemC-based test bench. The results show that compared with the negative first for Torus networks （NF-T） algorithm, the proposed algorithm can achieve better performance in terms of network latency and throughput under different traffic patterns. It also shows that a routing algorithm with load balance for virtual channels can significantly improve the network performance furthermore.展开更多
To improve the scalability and reduce the implementation complexity of Mesh and Mesh-like networks, the semi-diagonal Torus （SD-Torus） network, a regular and symmetrical intercormection network is proposed. The SD-T...To improve the scalability and reduce the implementation complexity of Mesh and Mesh-like networks, the semi-diagonal Torus （SD-Torus） network, a regular and symmetrical intercormection network is proposed. The SD-Torus network is a combination of a typical 2D-Torus network with two extra diagonal links from northwest to southeast direction for each node. The topological properties of SD-Torus networks are discussed, and a load balanced routing algorithm for SD-Torus is presented. System-C based simulation result shows that, compared with diagonal Mesh （DMesh）, diagonal Torus （DTorus） and XMesh networks, the SD-Torus network can achieve high performance with a lower network cost. It makes the SD-Torus network a powerful candidate for the high performance interconnection networks.展开更多
文摘本文设计了一种新的深度神经网络(New Deep Neural Network,NDNN)模型,并将其应用到入侵检测系统中.NDNN以其突出的特征学习能力充分学习训练数据的特征,在输出层,NDNN通过Softmax分类器对网络攻击报文与正常报文数据进行识别和分类,检测异常报文与入侵攻击.实验通过对KDD Cup 99数据集进行仿真,实验结果表明本文设计的基于NDNN的入侵检测系统模型,进一步提高了入侵检测系统的精度,增强了网络的安全性.
基金supported by the National Natural Science Foundation of China (60976020)
文摘In order to improve the network performance furthermore, a routing algorithm for 2D-Torus is investigated from the standpoint of load balance for virtual channels. The 2D-Torus network is divided into two virtual networks and each physical channel is split into three virtual channels. A novel virtual channel allocation policy and a routing algorithm are proposed, in which traffic load is distributed to those three virtual channels in a more load-balanced manner by introducing a random parameter. Simulations of the proposed algorithm are developed with a SystemC-based test bench. The results show that compared with the negative first for Torus networks （NF-T） algorithm, the proposed algorithm can achieve better performance in terms of network latency and throughput under different traffic patterns. It also shows that a routing algorithm with load balance for virtual channels can significantly improve the network performance furthermore.
基金supported by the National Natural Science Foundation of China (60976020)
文摘To improve the scalability and reduce the implementation complexity of Mesh and Mesh-like networks, the semi-diagonal Torus （SD-Torus） network, a regular and symmetrical intercormection network is proposed. The SD-Torus network is a combination of a typical 2D-Torus network with two extra diagonal links from northwest to southeast direction for each node. The topological properties of SD-Torus networks are discussed, and a load balanced routing algorithm for SD-Torus is presented. System-C based simulation result shows that, compared with diagonal Mesh （DMesh）, diagonal Torus （DTorus） and XMesh networks, the SD-Torus network can achieve high performance with a lower network cost. It makes the SD-Torus network a powerful candidate for the high performance interconnection networks.
