Hybrid Classifier for Crowd Anomaly Detection with Bernoulli Map Evaluation

Android has emerged as the dominant operating system over the past decade, enjoying widespread usage. However, its popularity has also made it an attractive target for malware proliferation. One reason behind this vulnerability is Android’s acceptance of app installations from unauthorized sources, which gives a window to malicious code writers to disseminate applications malicious in nature which in turn compromise the Android device. To tackle this issue, much research in the literature tried to develop malware analysis and detection systems that employ both static and dynamic analysis techniques. These approaches aim to identify and mitigate malware threats. However, existing research in this field still faces challenges in terms of efficiently and accurately detecting malware. Moreover, achieving accurate malware detection often requires significant resources, which can be a concern on resource-limited mobile devices. Despite advancements, the current state of research struggles to strike a balance between detection accuracy and resource utilization, particularly in the context of mobile devices with limited capabilities. In the current research paper, we present a new methodology for constructing a hybrid malware detection model that operates independently without relying on a pre-existing Machine Learning classifier. Our proposed model operates by leveraging the concept of feature-pairs, where pairs of features that exhibit strong correlations are identified. By calculating the feature-pair scores, we determine whether an application is benign or malicious based on these scores. This self-sufficient model eliminates the need for an external classifier, offering a unique approach to malware detection.

Due to the rapid growth of network technology, huge volume and distinct data sent via networks is expanding constantly. The situation shows how complex and dense cyber attacks and hazards are developing. Due to the rapid advancement in network density, cyber security specialists find it difficult to monitor all network activity. Due to frequent and sophisticated cyber attacks, it is becoming more challenging to detect and identify abnormalities in network events. The use of deep learning provides a variety of tools and strategies for automated cyber-attack detection as well as quick attack-type prediction as well as evaluation. This work introduces a novel cyber-attack detection and mitigation process under the following phases including preprocessing, feature extraction via the Map Reduce framework that handles the big data, feature selection, attack detection and mitigation. The Improved Normalisation process is achieved on the preprocessing phase. The work is examined from a big data perspective; hence Map Reduce framework is utilised for this. As a result, the framework will manage the feature extraction process, where features including statistical features, raw features, improved correlation-based features, and info gain-based features will be extracted. Following feature extraction, the Improved Recursive Feature Elimination procedure is processed that selects the relevant features. The hybrid detection model, which combines Recurrent Neural Networks (RNN) Deep and Belief Networks (DBN) is used to detect the attacks. Once an attack has been detected, the attacker must be mitigated. To accomplish this, an improved BAIT-based mitigation procedure is used. The two datasets used in this work are, namely, Intrusion Detection Systems (IDS) 2018 Intrusion CSVs (CSE-CIC-IDS2018) and UNSW_NB15. Finally, the suggested model and the alternative methods are contrasted using a variety of measures such as accuracy, sensitivity, specificity, precision, FDR, FNR and FPR.

With the continuous development of cloud technology, the services carried by the cloud platform continue to increase, which brings higher requirements to the network link resource planning and operation management. Establishing a link traffic prediction model for the actual needs of the network enables network operators to The rapid decision-making of resource scheduling is more effective with the assistance of advanced control and forwarding technologies such as SDN. Existing prediction methods do not consider the deep fine-grained features of current network traffic that are more diverse and more complex. By improving the LSTM model, this paper is oriented towards scientific research. To meet the management requirements of private networks, a new link traffic prediction model is proposed. The AE-stacked hybrid LSTM model consists of autoencoder AE, bidirectional LSTM model, unidirectional LSTM model and fully connected layer. The ability to extract traffic features can better mine the dependencies between data features at different times. The experimental results show that the prediction results of this model are in line with the real trend of traffic changes, and the residual between the predicted value and the observed value is small. It can better fit the existing traffic of the scientific research network.

Segmentation ability map: Interpret deep features for medical image segmentation.

In this study, a novel framework using multiple novelty detection filters is developed to learn a model of the normality of a robot’s visual perception, which is called multichannel novelty detection. Subsequently, the acquired model was used to highlight dissimilar perceptions when the robot explored an environment. The main purpose of fusing multiple novelty filters is that each novelty filter performs well in detecting specific types of novelties; therefore, a new framework is proposed that demonstrates a new way to combine multiple different purposed novelty detection filters together in order to yield an overall more robust novelty status on the visual features. To develop a multichannel novelty detection system, expectation- and appearance-based novelty detection models were used in this study. To become experts in detecting different types of novelty using these models, different features from the input image were extracted as inputs for the models. The expectation-based novelty detection model uses the MobileNetV2 deep network to extract the deep features of the input image, which is subsequently used to learn a sequential and temporal model of normality to detect novelty. By contrast, the appearance-based novelty detection model uses speeded up robust features (SURF), which provide more region-focused features within the input image, to identify whether a specific region of the image is novel. The proposed multichannel novelty detection system is a completely online and real-time approach that is very important for mobile robotics applications. The proposed framework was tested in three novel environments, and it was reported that the proposed multichannel novelty detection system performs better than expectation-based and appearance-based novelty filters separately. Statistically, in the three novel environments, the Matthews correlation coefficients are reported to be 0.94, 0.97, and 0.93, and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">F</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> scores are reported to be 0.95, 0.97, and 0.93, respectively, which proves that and can be concluded as almost perfect statistically.

Deep learning now plays an important role in solving complex problems in computer vision fields. The highly challenging high-resolution remote sensing image scene classification problem can also be solved using deep learning methods. The most commonly used method of deep learning is the convolutional neural network model. In this letter, based on deep learning, a combined model named Inception-long short-term memory (LSTM) is proposed. First, we combine the deep learning feature extracted from the pretrained Inception-V3 model with a hand-crafted feature: the GIST feature. The different features are then combined and input into the batch normalization (BN) layer. Second, the BN layer plays the role of the bridge to combine the InceptionV3 model with the LSTM model, which features a softmax classifier. The LSTM model is used to analyze the features and classify the different high-resolution remote sensing scene images. The proposed model, as a whole, can be uniformly trained. Three different datasets-the NWPU-RESISC45 dataset, the UC Merced dataset, and the SIRI-WHU dataset-were used to verify the effectiveness of the proposed model. The results show that the proposed Inception-LSTM model shows an outstanding performance in the scene classification task.

Distributed Denial of Service (DDOS) attack is one of the most common network attacks. DDoS attacks are becoming more and more diverse, which makes it difficult for some DDoS attack detection methods based on single network flow characteristics to detect various types of DDoS attacks, while the detection methods of multi-feature DDoS attacks have a certain lag due to the complexity of the algorithm. Therefore, it is necessary and urgent to monitor the trend of traffic change and identify DDoS attacks timely and accurately. In this paper, a method of DDoS attack detection based on deep belief network feature extraction and LSTM model is proposed. This method uses deep belief network to extract the features of IP packets, and identifies DDoS attacks based on LSTM model. This scheme is suitable for DDoS attack detection technology. The model can accurately predict the trend of normal network traffic, identify the anomalies caused by DDoS attacks, and apply to solve more detection methods about DDoS attacks in the future.

The process of 2D to 3D involves depth map estimation of input image and stereo pair generation using that depth image. Fully automatic techniques are not able to give very high quality output and methods which involve human operators are costly and time consuming. We propose a semiautomatic technique which incorporates monocular depth cues if these are not present in the input image. We propose a method which first includes defocusing factor in the input image semi-automatically and then estimates depth from focus cue. We also used motion parallax as another monocular depth cue and evaluate the performance of final depth map. Gradient Magnitude Similarity Deviation which is a highly efficient perceptual quality index is used for the performance evaluation of final depth map. GMSD score is lowest in case of final depth map which shows highest quality. Finally, stereo pair is generated using original image and its depth map.

Deep capsule networks have more capsule layers, which makes their performance better on complex images. However, with the increase of layers, overfitting will become more serious. Image reconstruction is an effective regularization method for capsule networks. To improve it, we propose an adversarial decoder that introduces the generative adversarial network framework into the reconstruction process to implement learnable reconstruction losses. This architecture consists of three parts: a deep capsule network, a decoder, and a discriminator. The deep capsule network extracts feature capsules from input images, which are then reconstructed by the decoder. The discriminator is the learnable reconstruction loss function that evaluates the similarity between reconstructed images and input images. Minimizing this learnable reconstruction loss and mean square error of images provides a regularization effect for the deep capsule network. Experimental results show that our models have a competitive performance of regularization on CIFAR10, CIFAR100, and FMNIST.

The corrugated layer structure bismuth has been successfully tailored into negative thermal expansion along c axis by size effect. Pair distribution function and extended X-ray absorption fine structure are combined to reveal the local structural distortion for nanosized bismuth. The comprehensive method to identify the local structure of nanomaterials can benefit the regulating and controlling of thermal expansion in nanodivices.

Abstract: This project presents a Content-Based Image Retrieval (CBIR) system that utilizes deep learning to improve the efficiency and accuracy of image similarity search. The system leverages a pre-trained ResNet-50 convolutional neural network, repurposed as a deep feature extractor by removing its final classification layers. Input images are first pre- processed and then passed through the network to extract high-dimensional feature vec- tors that capture rich visual semantics. These deep features are compared using cosine similarity to identify visually similar images. The system supports real-time image up- loads and retrieval by matching queries against a pre computed dataset of image fea- tures, enabling fast and responsive search capabilities. By replacing traditional hand- crafted features with deep feature representations, the system achieves significantly higher retrieval accuracy and robustness across various image types and domains. This approach demonstrates strong potential for practical deployment in areas such as digital asset management, visual search engines, and e-commerce product discovery, where visual similarity plays a critical role. Overall, the integration of deep learning into CBIR systems represents a significant advancement in the field of image search and retrieval. Keywords: CBIR, Deep Learning, ResNet-50, Feature Extraction, Image Simi- larity

In low-dose computed tomography (LDCT) denoising tasks, it is often difficult to balance edge/detail preservation and noise/artifact reduction. To solve this problem, we propose a dual convolutional neural network (CNN) based on edge feature extraction (Ed-DuCNN) for LDCT. Ed-DuCNN consists of two branches. One branch is the edge feature extraction subnet (Edge_Net) that can fully extract the edge details in the image. The other branch is the feature fusion subnet (Fusion_Net) that introduces an attention mechanism to fuse edge features and noisy image features. Specifically, first, shallow edge-specific detail features are extracted by trainable Sobel convolutional blocks and then are integrated into Edge_Net together with the LDCT images to obtain deep edge detail features. Finally, the input image, shallow edge detail, and deep edge detail features are fused in Fusion_Net to generate the final denoised image. The experimental results show that the proposed Ed-DuCNN can achieve competitive performance in terms of quantitative metrics and visual perceptual quality compared with that of state-of-the-art methods.

Based on the dust concentration data and meteorological environment data monitored at the open-pit mine site, the characteristics of dust concentration and the influence of temperature, humidity, wind speed, air pressure and other meteorological conditions on dust concentration were analyzed, and the causes of the change of dust concentration were clarified. Meanwhile, a dust concentration prediction model based on LSTM neural network is established. The results show that the dust concentration of the open-pit mine is high in March, November and the whole winter, and it is low in summer and autumn. The daily variation of humidity and temperature in different seasons showed the trend of “herringbone” and “inverted herringbone”, respectively. In addition, the wind speed was the highest in spring and the air pressure distribution was uniform, which basically maintained at 86–88 kPa. The peak humidity gradually deviates with each month and is obviously affected by seasonality. The higher the humidity, the lower the temperature and the higher the concentration of dust. In different seasons, the wind speed is the highest around 20:00 at night, and the dust is easy to disperse. The R2 values of PM2.5, PM10 and TSP concentrations predicted by LSTM model are 0.88, 0.87 and 0.87, respectively, which were smaller than the MAE, MAPE and RMSE values of other prediction models, and the prediction effect was better with lower error. The research results can provide a theoretical basis for dust distribution law, concentration prediction and dust removal measures of main dust sources in open-pit mines.

Different deep convolution neural network (DCNN) models have been proposed for wafer map pattern identification and classification tasks in previous studies. However, factors such as the effect of input image resolution on the classification performance of the proposed models and class imbalance in the training set after splitting the data into training and test sets have not been considered in the previous studies. This study proposes a DCNN model with residual blocks, called Opt-ResDCNN model, for wafer map defect pattern identification and classification by considering 26 * 26 input image resolutions and class imbalance issues during the model training. The proposed model is compared with the previously published defect pattern recognition and classification models in terms of accuracy, precision, recall, and F1 score for 26 * 26 input image size. Using a publicly available wafer map dataset (WM-811K), the proposed method can obtain an average accuracy, precision, recall, and F1 score results of 99.672%, 99.664%, 99.695%, 99.692%, respectively for the 26 * 26 input image resolution.

Micro-cracks detection of solar cells surface via combining short-term and long-term deep features