State Grid Corporation of China has researched and released dual-mode communication technology standards based on High Speed Power Line Communication (HPLC) and Radio Frequency (RF) for collecting electricity consumption data in low-voltage distribution metering networks. Due to occasional unknown changes, the topology of the low-voltage distribution metering network has changed, and some nodes cannot even be used. In this paper, we utilized attention mechanisms to accurately locate key features and combined them with Long Short-Term Memory (LSTM) networks to reveal long-term dependencies in the data. Based on this innovative approach, we propose the AT-LSTM model, which is an efficient detection method designed specifically for time series data Unlike traditional anomaly detection algorithms, it can detect and locate node channel anomalies in dual-mode communication networks without any relevant domain knowledge, combined with identified communication network topology, and predict dual-mode channel state information at any given time point based on historical data. Realize topology change recognition and communication anomaly detection when the prediction error follows an approximate Gaussian distribution. The effectiveness of the model proposed in this paper was demonstrated through evaluation using synthetic data extracted from low-voltage distribution networks.

Video question answering (Video-QA) is a subject undergoing intense study in Artificial Intelligence, which is one of the tasks which can evaluate such AI abilities. In this paper, we propose a Modality Attention Fusion framework with Hybrid Multi-head Self-attention (MAF-HMS). MAF-HMS focuses on the task of answering multiple-choice questions regarding a video-subtitle-QA representation by fusion of attention and self-attention between each modality. We use BERT to extract text features, and use Faster R-CNN to ex-tract visual features to provide a useful input representation for our model to answer questions. In addition, we have constructed a Modality Attention Fusion (MAF) framework for the attention fusion matrix from different modalities (video, subtitles, QA), and use a Hybrid Multi-headed Self-attention (HMS) to further determine the correct answer. Experiments on three separate scene datasets show our overall model outperforms the baseline methods by a large margin. Finally, we conducted extensive ablation studies to verify the various components of the network and demonstrate the effectiveness and advantages of our method over existing methods through question type and required modality experimental results.

Social media allows users to express opinions in multiple modalities such as text, pictures, and short-videos. Multi-modal sentiment detection can more effectively predict the emotional tendencies expressed by users. Therefore, multi-modal sentiment detection has received extensive attention in recent years. Current works consider utterances from videos as independent modal, ignoring the effective interaction among diffence modalities of a video. To tackle these challenges, we propose transformer-based interactive multi-modal attention network to investigate multi-modal paired attention between multiple modalities and utterances for video sentiment detection. Specifically, we first take a series of utterances as input and use three separate transformer encoders to capture the utterances-level features of each modality. Subsequently, we introduced multimodal paired attention mechanisms to learn the cross-modality information between multiple modalities and utterances. Finally, we inject the cross-modality information into the multi-headed self-attention layer for making final emotion and sentiment classification. Our solutions outperform baseline models on three multi-modal datasets.

Malicious traffic detection is one of the most important parts of cyber security. The approaches of using the flow as the detection object are recognized as effective. Benefiting from the development of deep learning techniques, raw traffic can be directly used as a feature to detect malicious traffic. Most existing work usually converts raw traffic into images or long sequences to express a flow and then uses deep learning technology to extract features and classify them, but the generated features contain much redundant or even useless information, especially for encrypted traffic. The packet header field contains most of the packet characteristics except the payload content, and it is also an important element of the flow. In this paper, we only use the fields of the packet header in the raw traffic to construct the characteristic representation of the traffic and propose a novel flow-vector generation approach for malicious traffic detection. The preprocessed header fields are embedded as field vectors, and then a two-layer attention network is used to progressively generate the packet vectors and the flow vector containing context information. The flow vector is regarded as the abstraction of the raw traffic and is used to classify. The experiment results illustrate that the accuracy rate can reach up to 99.48% in the binary classification task and the average of AUC-ROC can reach 0.9988 in the multi-classification task. (C) 2022 Elsevier Inc. All rights reserved.

Network attacks are one of the main threats to the stable operation of smart grid equipment. As a real-time monitoring system to prevent network attacks, intrusion detection is widely used in smart grid protection. However, the massive data in the network transmission process contains a large number of redundant and irrelevant features, which makes it difficult for the intrusion detection system to process in time and reduce the efficiency. Feature selection is a method to solve this kind of problem. It can improve the speed of intrusion detection by filtering the characteristics of massive data. Therefore, a hybrid feature selection algorithm which combines information gain and genetic search to improve the work efficiency of intrusion detection systems is proposed. The algorithm is mainly divided into three parts. Firstly, the information gain value of all features is calculated by using information gain, according to which all features are ordered, and the ordered features is ranked according to an exponential increase strategy; secondly, the ranked features is used to guide the genetic algorithm search process, and a new fitness function can be used to control the search direction of genetic algorithm; finally, a classification algorithm is used to test the dataset after feature selection. In experiments, by comparing with other feature selection algorithms on 5 sets of high-dimensional UCI datasets, it is concluded that the IGExpGA proposed in this paper significantly improves the detection rate and detection speed. More importantly, in the KDD1998 network data, the algorithm proposed improves the detection rate to 98.8%, which is significantly better than other algorithms.

Gated Recurrent Unit (GRU) is a variant of a recurrent neural network, just like an LSTM network. Compared with RNN, the two networks have higher accuracy in processing sequence problems, and both of them have been proven to be effective in varieties of machine learning tasks such as natural language processing, text classification and speech recognition. In addition, the network unit structure of the GRU is simpler than the LSTM unit structure, which is more conducive to the training of the model. NSL-KDD datasets, which is the replacement of KDD cup 99, is still one of the datasets for measuring the effectiveness of intrusion detection models. In order to reduce the feature data dimension and combine the prior knowledge of computer network, a GRU intrusion detection method based on salient features (SF-GRU) is proposed. SF-GRU selects the distinctive features of response for different intrusion forms, and uses GRU network to identify the selected features to improve the efficiency of model detection. The experimental results show that compared with the traditional deep learning method, this proposal has higher accuracy and computational efficiency.