In imbalanced network traffic, malicious cyber-attacks can often hide in large amounts of normal data. It exhibits a high degree of stealth and obfuscation in cyberspace, making it difficult for Network Intrusion Detection System(NIDS) to ensure the accuracy and timeliness of detection. This paper researches machine learning and deep learning for intrusion detection in imbalanced network traffic. It proposes a novel Difficult Set Sampling Technique(DSSTE) algorithm to tackle the class imbalance problem. First, use the Edited Nearest Neighbor(ENN) algorithm to divide the imbalanced training set into the difficult set and the easy set. Next, use the KMeans algorithm to compress the majority samples in the difficult set to reduce the majority. Zoom in and out the minority samples’ continuous attributes in the difficult set synthesize new samples to increase the minority number. Finally, the easy set, the compressed set of majority in the difficult, and the minority in the difficult set are combined with its augmentation samples to make up a new training set. The algorithm reduces the imbalance of the original training set and provides targeted data augment for the minority class that needs to learn. It enables the classifier to learn the differences in the training stage better and improve classification performance. To verify the proposed method, we conduct experiments on the classic intrusion dataset NSL-KDD and the newer and comprehensive intrusion dataset CSE-CIC-IDS2018. We use classical classification models: random forest(RF), Support Vector Machine(SVM), XGBoost, Long and Short-term Memory(LSTM), AlexNet, Mini-VGGNet. We compare the other 24 methods; the experimental results demonstrate that our proposed DSSTE algorithm outperforms the other methods.

In the context of the proliferated evolution of network service types and the expeditious augmentation of network resource deployment, the requisition for copious labeled datasets to facilitate superior performance in traffic classification methods, particularly those hinging on deep learning, is imperative. Nonetheless, the procurement and annotation of such extensive datasets necessitate considerable temporal and human resource investments. In response to this predicament, this work introduces a methodology, termed MTEFU, leveraging a deep learning model-based multi-task learning algorithm, strategically designed to mitigate the reliance on substantial labeled training samples. Multiple classification tasks, encompassing duration, bandwidth size, and business traffic category, are incorporated, with a shared parameter strategy implemented amongst tasks to assure the transference of information across disparate tasks. Employing CNN, SAE, GRU, and LSTM as multi-task learning classification models, training validation and experimental testing were conducted on the QUIC dataset. A comparative analysis with single-task and ensemble learning methods reveals that, in the context of predicting network traffic types, the accuracy derived from the multi-task learning strategy, even with a mere 150 labeled samples, can emulate the 94.67% accuracy achieved through single-task learning with a fully labeled dataset of 6139 samples.

Based on the design idea of future network, this paper analyzes the network security data sampling and anomaly prediction in future network. Through game theory, it is determined that data sampling is performed on some important nodes in the future network. Deep learning methods are used on the selected nodes to collect data and analyze the characteristics of the network data. Then, through offline and real-time analyses, network security abnormal events are predicted in the future network. With the comparison of various algorithms and the adjustment of hyperparameters, the data characteristics and classification algorithms corresponding to different network security attacks are found. We have carried out experiments on the public dataset, and the experiment proves the effectiveness of the method. It can provide reference for the management strategy of the switch node or the host node by the future network controller.

Traversable regions identification technology plays a crucial role in ensuring safe driving for unmanned ground vehicles in off-road environments. However, the unstructured terrain makes it challenging to identify traversable regions. To enhance the safety of off-road driving, a LiDAR-based traversable regions identification method is proposed in this paper. Firstly, a deep learning-based neural network is used to segment the traversable regions, obstacles, and vegetation. Next, an improved Gaussian Process(GP)-based modeling method is designed to model the traversable regions with a leading speed, and the obstacle point clouds are refined with a composite filter. Finally, field experiments have demonstrated that our proposed scheme outperforms existing state-of-the-art (SOTA) traditional and deep-learning-based methods in accurately identifying both road regions and obstacles, with precision improvements of up to 14% and recall improvements of up to 9%.

The path planning of flexible needles in robotic puncturing presents challenges such as limited model accuracy and poor real-time performance, which affect both efficiency and accuracy in complex medical scenarios. To address these issues, this paper proposes a deep reinforcement learning-based path planning method for flexible needles in robotic puncturing. Firstly, we introduce a unicycle model to describe needle motion and design a hierarchical model to simulate layered tissue interactions with the needle. The forces exerted by tissues at different positions on the flexible needle are considered, achieving a combination of kinematic and mechanical models. Secondly, a deep reinforcement learning framework is built, integrating obstacle avoidance and target attraction to optimize path planning. The design of state features, the action space, and the reward function is tailored to enhance the model’s decision-making capabilities. Moreover, we incorporate a retraction mechanism to bolster the system’s adaptability and robustness in the dynamic context of surgical procedures. Finally, laparotomy simulation results validate the proposed method’s effectiveness and generalizability, demonstrating its superiority over current state-of-the-art techniques in robotic puncturing.

Based on comprehensive consideration of stochastic factors and the system performance, we presented an integrated queuing network model (IQNM) to effectively optimize the facility layout of the multi-floor flow shop (MFFS). In this paper, the arrival interval and service times of the workstations are described by the two-moment approximation method, and an iterative algorithm is proposed for estimating the system performance, including the mean throughput, work-in-process (WIP) and sojourn time. In the design process, an optimization model with the throughput and cycle time constraint is established, while the minimize transportation cost is taken as the objective function. After several iterations between system performance and optimization objective, the facility layout problem (FLP) of the MFFS is solved, which takes the maximize throughput and minimize cycle time as the constraint condition. Finally, the effectiveness of the proposed method is verified by a series of numerical experiments. The average error of system throughput is 2.11%. And the sojourn time errors are less than 5% when SCV = 1.5. Therefore, this research can provide relevant theoretical support for the facility layout of the MFFS.