No abstract

No abstract

Currently, Low-Rate Denial of Service (LDoS) attacks are one of the main threats faced by Software-Defined Wireless Sensor Networks (SDWSNs). This type of attack uses a lot of low-rate requests to occupy network resources and hard to detect. An efficient detection method has been proposed for LDoS attacks with the features of small signals. The non-smooth small signals generated by LDoS attacks are analyzed employing the time–frequency analysis method based on Hilbert–Huang Transform (HHT). In this paper, redundant and similar Intrinsic Mode Functions (IMFs) are removed from standard HHT to save computational resources and to eliminate modal mixing. The compressed HHT transformed one-dimensional dataflow features into two-dimensional temporal–spectral features, which are further input into a Convolutional Neural Network (CNN) to detect LDoS attacks. To evaluate the detection performance of the method, various LDoS attacks are simulated in the Network Simulator-3 (NS-3) experimental environment. The experimental results show that the method has 99.8% detection accuracy for complex and diverse LDoS attacks.

When training a large-scale knowledge graph embedding (KGE) model with multiple graphics processing units (GPUs), the partition-based method is necessary for parallel training. However, existing partition-based training methods suffer from low GPU utilization and high input/output (IO) overhead between the memory and disk. For a high IO overhead between the disk and memory problem, we optimized the twice partitioning with fine-grained GPU scheduling to reduce the IO overhead between the CPU memory and disk. For low GPU utilization caused by the GPU load imbalance problem, we proposed balanced partitioning and dynamic scheduling methods to accelerate the training speed in different cases. With the above methods, we proposed fine-grained partitioning KGE, an efficient KGE training framework with multiple GPUs. We conducted experiments on some benchmarks of the knowledge graph, and the results show that our method achieves speedup compared to existing framework on the training of KGE.

No abstract