Feature-based Methods Research Articles

Malware2ATT&CK: A sophisticated model for mapping malware to ATT&CK techniques

MITER Corporation presents the ATT&CK Matrix, which maps malware behavior to different tactics, techniques, and procedures (TTPs) providing a comprehensive view to clarify the inner mechanism of malware more accurately. However, manual mapping methods are time-consuming, while rule-based feature-based mapping methods often under- or misreport many attacks. Inspired by the successful application of image multi-label classification techniques, we propose a method called Malware2ATT&CK to automatically map malware to ATT&CK techniques. The method applies pre-trained models to extract features from the two pieces of information — the static analysis information of assembly instruction and API calls from malware. The malicious techniques are identified by a multi-label classifier based on the graph neural network and knowledge graph. In the experiments over two test sets, Malware2ATT&CK shows excellent performance achieving an average F1 score of 83.6% in the technology mapping task. Further evaluation indicates that the high accuracy of the prediction is due to our ability to accurately capture the correlation between behaviors.

RockS2Net: Rock image classification via a spatial localization siamese network

The acquisition of rock property information is at the core of regional geological survey and mineral exploration, but hand-crafted feature-based methods are heavily influenced by human prior knowledge and have limited transferability. End-to-end deep learning techniques, exemplified by convolutional neural networks (CNNs), have attained significant accomplishments in the domain of image classification. However, previous end-to-end CNN-based methods are hard to focus on image critical areas, and they also cannot make full use of global alignment dependency of the rocks. In this paper, rock image classification via a RockS2Net is proposed. The RockS2Net framework adopts a Siamese architecture, characterized by two branches that share parameters, enabling the efficient extraction of both global and local features. This approach facilitates the extraction of global features from entire images and focuses on critical areas to extract local features. The architecture of spatial transformer network (STN) is introduced to transform microscopic images of rock sections to their critical areas. By fusing local features and global features, the properties obtained from microscopic images of rock sections can be more accurately predicted. To test the robust generalizability of the proposed method, the constructed CHN-Rock images dataset is used for experiments and evaluation. Experimental results show that the accuracy of the proposed RockS2Net on the CHN-Rock image dataset is 2–3% higher than that of other rock image classification networks.

Optimizing AUV Navigation Using Factor Graphs with Side-Scan Sonar Integration

For seabed mapping, the prevalence of autonomous underwater vehicles (AUVs) employing side-scan sonar (SSS) necessitates robust navigation solutions. However, the positioning errors of traditional strapdown inertial navigation system (SINS) and Doppler velocity log (DVL) systems accumulated significantly, further exacerbated by DVL’s susceptibility to failure in complex underwater conditions. This research proposes an integrated navigation approach that utilizes factor graph optimization (FGO) along with an improved pre-integration technique integrating SSS-derived position measurements. Firstly, the reliability of SSS image registration in the presence of strong noise and feature-poor environments is improved by replacing the feature-based methods with a Fourier-based method. Moreover, the high-precision inertial measurement unit (IMU) pre-integration method could correct the heading errors of SINS significantly by considering the Earth’s rotation. Finally, the AUV’s marine experimental results demonstrated that the proposed integration method not only offers improved SSS image registration and corrects initial heading discrepancies but also delivers greater system stability, particularly in instances of DVL data loss.

Offline Handwritten Signature Identification based on Hybrid Features and Proposed Deep Model

Handwritten signature identification is the process of identifying the true identity of an individual by analyzing their signature. This is an important task in applications such as financial transactions, legal documents, and biometric systems. Various techniques have been developed for signature identification, including feature-based methods and machine learning-based methods. This paper proposes an authentic signature identification method based on integrating static (off-line) signature data and proposed deep-based model, this is done by fused three types of signature features, Linear Discriminant Analysis (LDA) as appearance-based features, Fast Fourier Transform (FFT) as frequency-features, and Gray-Level Co-occurrence Matrix (GLCM) as texture-features. Then, the fused features are inputted to the proposed deep-based model of 25 layers for identifying each person. For experiments, we have used three datasets: Our own private collected dataset, called SigArab, and two public datasets called SigComp11 and CEDAR respectively. The proposed deep model achieves 99.23%, 100%, 100% accuracy on SigArab, CEDAR and SigComp2011 datasets.

HSCNet++: Hierarchical Scene Coordinate Classification and Regression for Visual Localization with Transformer

Visual localization is critical to many applications in computer vision and robotics. To address single-image RGB localization, state-of-the-art feature-based methods match local descriptors between a query image and a pre-built 3D model. Recently, deep neural networks have been exploited to regress the mapping between raw pixels and 3D coordinates in the scene, and thus the matching is implicitly performed by the forward pass through the network. However, in a large and ambiguous environment, learning such a regression task directly can be difficult for a single network. In this work, we present a new hierarchical scene coordinate network to predict pixel scene coordinates in a coarse-to-fine manner from a single RGB image. The proposed method, which is an extension of HSCNet, allows us to train compact models which scale robustly to large environments. It sets a new state-of-the-art for single-image localization on the 7-Scenes, 12-Scenes, Cambridge Landmarks datasets, and the combined indoor scenes.

Channel-level Matching Knowledge Distillation for object detectors via MSE

Knowledge distillation (KD) has been widely used in different tasks as a practical model compression technique. Due to the poor performance of directly using Mean Square Error (MSE) between the intermediate features of the teacher and student, most feature-based detector distillation methods are primarily concerned with proposing diverse attention mechanisms and employing MSE to guide the student in learning critical information. However, the significance of MSE in detector distillation is often overlooked. To enhance the distillation performance of MSE and provide a novel perspective, we propose channel-level Matching Knowledge Distillation via MSE (MKD). The idea of MKD is simple but effective: based on the match or mismatch of student and teacher channel features in Feature Pyramid Networks (FPN), we divide channels at the same location in the FPN layers of the student and teacher into match channels and mismatch channels and then assign different weights to them to calculate the MSE-based distillation loss, which forces the student to distinguish the teacher’s channels and imitate the features of the corresponding channels to achieve better performance. We conduct extensive experiments on MS COCO and PASCAL VOC benchmarks to prove that MKD is superior to state-of-the-art distillation methods. For example, with a powerful RetinaNet-ResNeXt101 detector as the teacher, ResNet-50 based RetinaNet and RepPoints with our distillation method achieve 40.9% and 42.2% mAP on COCO2017, which are 3.5% and 3.6% higher than the baseline, respectively.

Proposal of dental demineralization diagnosis with OCT echo based on multiscale entropy analysis.

Optical coherence tomography (OCT) has been widely used for the diagnosis of dental demineralization. Most methods rely on extracting optical features from OCT echoes for evaluation or diagnosis. However, due to the diversity of biological samples and the complexity of tissues, the separability and robustness of extracted optical features are inadequate, resulting in a low diagnostic efficiency. Given the widespread utilization of entropy analysis in examining signals from biological tissues, we introduce a dental demineralization diagnosis method using OCT echoes, employing multiscale entropy analysis. Three multiscale entropy analysis methods were used to extract features from the OCT one-dimensional echo signal of normal and demineralized teeth, and a probabilistic neural network (PNN) was used for dental demineralization diagnosis. By comparing diagnostic efficiency, diagnostic speed, and parameter optimization dependency, the multiscale dispersion entropy-PNN (MDE-PNN) method was found to have comprehensive advantages in dental demineralization diagnosis with a diagnostic efficiency of 0.9397. Compared with optical feature-based dental demineralization diagnosis methods, the entropy features-based analysis had better feature separability and higher diagnostic efficiency, and showed its potential in dental demineralization diagnosis with OCT.

Comparison of feature-based sentence ranking methods for extractive summarization of Turkish news texts

Comparison of feature-based sentence ranking methods for extractive summarization of Turkish news texts

TransMatch: A Transformer-Based Multilevel Dual-Stream Feature Matching Network for Unsupervised Deformable Image Registration.

Feature matching, which refers to establishing the correspondence of regions between two images (usually voxel features), is a crucial prerequisite of feature-based registration. For deformable image registration tasks, traditional feature-based registration methods typically use an iterative matching strategy for interest region matching, where feature selection and matching are explicit, but specific feature selection schemes are often useful in solving application-specific problems and require several minutes for each registration. In the past few years, the feasibility of learning-based methods, such as VoxelMorph and TransMorph, has been proven, and their performance has been shown to be competitive compared to traditional methods. However, these methods are usually single-stream, where the two images to be registered are concatenated into a 2-channel whole, and then the deformation field is output directly. The transformation of image features into interimage matching relationships is implicit. In this paper, we propose a novel end-to-end dual-stream unsupervised framework, named TransMatch, where each image is fed into a separate stream branch, and each branch performs feature extraction independently. Then, we implement explicit multilevel feature matching between image pairs via the query-key matching idea of the self-attention mechanism in the Transformer model. Comprehensive experiments are conducted on three 3D brain MR datasets, LPBA40, IXI, and OASIS, and the results show that the proposed method achieves state-of-the-art performance in several evaluation metrics compared to the commonly utilized registration methods, including SyN, NiftyReg, VoxelMorph, CycleMorph, ViT-V-Net, and TransMorph, demonstrating the effectiveness of our model in deformable medical image registration.

A novel time series classification for multivariate data using improved deep belief-recurrent neural network with optimal dynamic time warping

In the past ten years, data from time series extraction has attracted a lot of attention. Several methods have concentrated on classification problems, where the objective is to identify the labelling of a test period, given labelled training data. Feature-based and Instance-based methods are the two fundamental groups into which time series categorization methodologies may be divided. To categorize time series data, instance-based techniques use similarity data in a nearest-neighbor context. While methods in this category deliver reliable findings, their efficacy suffers when dealing with lengthy and noisy time series. Feature-based approaches, on the other together, extract characteristics to address the shortcomings of instance-based methods; nevertheless, these approaches use predetermined features and might not be effective in all classification issues. This paper seeks to introduce a novel deep learning-based Optimal Dynamic Time Warping (ODTW) paradigm for multimodal time’s series data categorization. This model covers several phases. At initial stage, the standard data is gathered from standard public source. Secondly, ODTW is proposed, where the parameters are optimized by Random Opposition Billiards-Inspired Optimization (RO-BIO) for extracting the most essential information. Finally, the classification is carried out through “Deep Belief Network (DBN) and Recurrent Neural Networks (RNN) termed as Deep Belief-RNN (DB-RNN)”. Finally, the extracted deep features are given to the optimized RNN for attaining the final classified results. The simulation results have resulted in superior classification performance in terms of standard performance measures.

From Scalp to Ear-EEG: A Generalizable Transfer Learning Model for Automatic Sleep Scoring in Older People.

Sleep monitoring has extensively utilized electroencephalogram (EEG) data collected from the scalp, yielding very large data repositories and well-trained analysis models. Yet, this wealth of data is lacking for emerging, less intrusive modalities, such as ear-EEG. The current study seeks to harness the abundance of open-source scalp EEG datasets by applying models pre-trained on data, either directly or with minimal fine-tuning; this is achieved in the context of effective sleep analysis from ear-EEG data that was recorded using a single in-ear electrode, referenced to the ipsilateral mastoid, and developed in-house as described in our previous work. Unlike previous studies, our research uniquely focuses on an older cohort (17 subjects aged 65-83, mean age 71.8 years, some with health conditions), and employs LightGBM for transfer learning, diverging from previous deep learning approaches. Results show that the initial accuracy of the pre-trained model on ear-EEG was 70.1%, but fine-tuning the model with ear-EEG data improved its classification accuracy to 73.7%. The fine-tuned model exhibited a statistically significant improvement (p < 0.05, dependent t-test) for 10 out of the 13 participants, as reflected by an enhanced average Cohen's kappa score (a statistical measure of inter-rater agreement for categorical items) of 0.639, indicating a stronger agreement between automated and expert classifications of sleep stages. Comparative SHAP value analysis revealed a shift in feature importance for the N3 sleep stage, underscoring the effectiveness of the fine-tuning process. Our findings underscore the potential of fine-tuning pre-trained scalp EEG models on ear-EEG data to enhance classification accuracy, particularly within an older population and using feature-based methods for transfer learning. This approach presents a promising avenue for ear-EEG analysis in sleep studies, offering new insights into the applicability of transfer learning across different populations and computational techniques. An enhanced ear-EEG method could be pivotal in remote monitoring settings, allowing for continuous, non-invasive sleep quality assessment in elderly patients with conditions like dementia or sleep apnea.

MSGL+: Fast and Reliable Model Selection-Inspired Graph Metric Learning

The problem of learning graph-based data structures from data has attracted considerable attention in the past decade. Different types of data can be used to infer the graph structure, such as graphical Lasso, which is learned from multiple graph signals or graph metric learning based on node features. However, most existing methods that use node features to learn the graph face difficulties when the label signals of the data are incomplete. In particular, the pair-wise distance metric learning problem becomes intractable as the dimensionality of the node features increases. To address this challenge, we propose a novel method called MSGL+. MSGL+ is inspired from model selection, leverages recent advancements in graph spectral signal processing (GSP), and offers several key innovations: (1) Polynomial Interpretation: We use a polynomial function of a certain order on the graph Laplacian to represent the inverse covariance matrix of the graph nodes to rigorously formulate an optimization problem. (2) Convex Formulation: We formulate a convex optimization objective with a cone constraint that optimizes the coefficients of the polynomial, which makes our approach efficient. (3) Linear Constraints: We convert the cone constraint of the objective to a set of linear ones to further ensure the efficiency of our method. (4) Optimization Objective: We explore the properties of these linear constraints within the optimization objective, avoiding sub-optimal results by the removal of the box constraints on the optimization variables, and successfully further reduce the number of variables compared to our preliminary work, MSGL. (5) Efficient Solution: We solve the objective using the efficient linear-program-based Frank–Wolfe algorithm. Application examples, including binary classification, multi-class classification, binary image denoising, and time-series analysis, demonstrate that MSGL+ achieves competitive accuracy performance with a significant speed advantage compared to existing graphical Lasso and feature-based graph learning methods.

Cross-attention learning enables real-time nonuniform rotational distortion correction in OCT.

Nonuniform rotational distortion (NURD) correction is vital for endoscopic optical coherence tomography (OCT) imaging and its functional extensions, such as angiography and elastography. Current NURD correction methods require time-consuming feature tracking/registration or cross-correlation calculations and thus sacrifice temporal resolution. Here we propose a cross-attention learning method for the NURD correction in OCT. Our method is inspired by the recent success of the self-attention mechanism in natural language processing and computer vision. By leveraging its ability to model long-range dependencies, we can directly obtain the spatial correlation between OCT A-lines at any distance, thus accelerating the NURD correction. We develop an end-to-end stacked cross-attention network and design three types of optimization constraints. We compare our method with two traditional feature-based methods and a CNN-based method on two publicly-available endoscopic OCT datasets. We further verify the NURD correction performance of our method on 3D stent reconstruction using a home-built endoscopic OCT system. Our method achieves a ∼3 × speedup to real time (26 ± 3 fps), and superior correction performance.

A critical analysis of image-based camera pose estimation techniques

Camera, and associated with its objects within the field of view, localization could benefit many computer vision fields, such as autonomous driving, robot navigation, and augmented reality (AR). After decades of progress, camera localization, also called camera pose estimation could compute the 6DoF pose of objects for a camera in a given image, with respect to different images in a sequence or formats. Structure feature-based localization methods have achieved great success when integrated with image matching or with a coordinate regression stage. Absolute and relative pose regression methods using transfer learning can support end-to-end localization to directly regress a camera pose but achieve a less accurate performance. Despite the rapid development of multiple branches in this area, a comprehensive, in-depth and comparative analysis is lacking to summarize, classify and compare, structure feature-based and regression-based camera localization methods. Existing surveys either focus on larger SLAM (Simultaneous Localization and Mapping) systems or on only part of the camera localization method, lack detailed comparisons and descriptions of the methods or datasets used, neural network designs such as loss designs, and input formats, etc. In this survey, we first introduce specific application areas and the evaluation metrics for camera localization pose according to different sub-tasks (learning-based 2D-2D task, 2D-3D task, and 3D-3D task). Then, we review common methods for structure feature-based camera pose estimation approaches, absolute pose regression and relative pose regression approaches by critically modelling the methods to inspire further improvements in their algorithms such as loss functions, and neural network structures. Furthermore, we summarize what are the popular datasets used for camera localization and compare the quantitative and qualitative results of these methods with detailed performance metrics. Finally, we discuss future research possibilities and applications.

A Review of Homography Estimation: Advances and Challenges

Images captured from different viewpoints or devices have often exhibited significant geometric and photometric differences due to factors such as environmental variations, camera technology differences, and shooting conditions’ instability. To address this problem, homography estimation has attracted much attention as a method to describe the geometric projection relationship between images. Researchers have proposed numerous homography estimation methods for single-source and multimodal images in the past decades. However, the comprehensive review and analysis of homography estimation methods, from feature-based to deep learning-based, is still lacking. Therefore, we provide a comprehensive overview of research advances in homography estimation methods. First, we provide a detailed introduction to homography estimation’s core principles and matrix representations. Then, we review homography estimation methods for single-source and multimodal images, from feature-based to deep learning-based methods. Specifically, we analyze traditional and learning-based methods for feature-based homography estimation methods in detail. For deep learning-based homography estimation methods, we explore supervised, unsupervised, and other methods in-depth. Subsequently, we specifically review several metrics used to evaluate these methods. After that, we analyze the relevant applications of homography estimation and show the broad application prospects of this technique. Finally, we discuss current challenges and future research directions, providing a reference for computer vision researchers and engineers.

Research on percussion-based bolt looseness monitoring under noise interference and insufficient samples

Loose bolts are the key cause of tower collapse. The maintenance of bolts in transmission towers remains at the stage of individual inspection, and it is urgent to develop practical bolt looseness monitoring methods for engineering applications. The percussion-based method has shown its full potential due to not relying on fixed signal generators and sensors. However, its performance can severely degrade under noise interference and data scarcity. To address this problem, the new voiceprint feature that resist noise interference and classifier that alleviate sample shortages are proposed in this study, namely all-pole group delay function (APGDF) and prototypical networks. The proposed new bolt torque monitoring method is composed of the two. Under the condition that each bolt torque has only 50 percussion sound samples and −6dB signal-to-noise ratio (SNR), the accuracy of the proposed method for single bolt looseness monitoring is 77.87 %, and the accuracy of multi-bolt looseness monitoring is 80.60 %, which is significantly better than existing feature-based methods. In addition, the monitoring of bolt looseness with highly few samples and the data dimensionality reduction analysis are carried out to demonstrate the superiority of this method. The percussion point needs to be close to the loose bolt to achieve the best monitoring effect, and the percussion intensity does not affect the monitoring effect. The proposed method is expected to become the fundamental mode of bolt looseness monitoring in engineering practice.

A Robust Multi-Local to Global with Outlier Filtering for Point Cloud Registration

As a prerequisite for many 3D visualization tasks, point cloud registration has a wide range of applications in 3D scene reconstruction, pose estimation, navigation, and remote sensing. However, due to the limited overlap of point clouds, the presence of noise and the incompleteness of the data, existing feature-based matching methods tend to produce higher outlier matches, thus reducing the quality of the registration. Therefore, the generation of reliable feature descriptors and the filtering of outliers become the key to solving these problems. To this end, we propose a multi-local-to-global registration (MLGR) method. First, in order to obtain reliable correspondences, we design a simple but effective network module named the local geometric network (LG-Net), which can generate discriminative feature descriptors to reduce the outlier matches by learning the local latent geometric information of the point cloud. In addition, we propose a multi-local-to-global registration strategy to further filter outlier matches. We compute the hypothetical transformation matrix from local patch matches. The point match evaluated as an inlier under multiple hypothetical transformations will receive a higher score, and low-scoring point matches will be rejected. Finally, our method is quite robust under different numbers of samples, as it does not require sampling a large number of correspondences to boost the performance. The numerous experiments on well-known public datasets, including KITTI, 3DMatch, and ModelNet, have proven the effectiveness and robustness of our method. Compared with the state of the art, our method has the lowest relative rotation error and relative translation error on the KITTI, and consistently leads in feature matching recall, inlier ratio, and registration recall on 3DMatch under different numbers of point correspondences, which proves the robustness of our method. In particular, the inlier ratio is significantly improved by 3.62% and 4.36% on 3DMatch and 3DLoMatch, respectively. In general, the performance of our method is more superior and robust than the current state of the art.

Markov-Driven Graph Convolutional Networks for Social Spammer Detection

With the growing popularity of social media, spammers unfairly overpower legitimate users with unwanted content to achieve their illegal purposes, which encourages research on spammer detection. The existing spammer detection methods can be characterized into feature-based and propagation-based detection. However, feature-based methods (e.g., GCN) cannot capture the user's following relations, while propagation-based methods cannot utilize the rich text features. To this end, we consider combining these two methods and propose an Adaptive Reward Markov Random Field(ARMRF) layer. ARMRF layer models three intuitions on user label relations and assign them different learnable rewards. Besides, we learn the reward weights by stacking the ARMRF layer on top of GCN for end-to-end training, and we call the stacked model ARMGCN. To further improve ARMGCNs expressive ability, we propose the Markov-Driven Graph Convolutional Network(MDGCN), which integrates conditional random fields(CRF) and ARMGCN. CRF establishes the label joint probability distribution conditioned features for learning user dependencies, and the distribution can be optimized by a variational EM algorithm. We extensively evaluate the proposed method on two real-world Twitter datasets, and the experimental results demonstrate that MDGCN outperforms the state-of-the-art baselines. In addition, the ARMRF layer can be integrated with existing detection methods to improve performance further.

Remote access trojan traffic early detection method based on Markov matrices and deep learning

Remote Access Trojan (RAT) allows the attacker to gain remote control of an infected system and steal data from it. Due to over-reliance on expert experience and statistical features, most feature-based RAT detection methods perform inefficiently and can only achieve high accuracy with network traffic collected over a long period of time. Byte-based RAT detection methods often use sequence truncation to process network traffic to meet the uniform-sized input requirements of convolutional neural networks (CNNs). However, the sequence truncation brings a negative impact on detection accuracy for some information loss. Towards effective and efficient RAT detection, we define the early stage from the damage degree caused on victim hosts and propose a new RAT traffic early detection method based on Markov matrices and deep learning (RATMD). In RATMD, the byte sequence of the TCP payloads of the TCP service packets collected in the early stage is represented by a byte transition probability matrix with a fixed size of 256 × 256, and the generated matrices are used to construct a detection model using a CNN architecture. Experiments are conducted on the network traffic of 58 benign applications and 61 RATs. Employing only the byte sequences derived from the TCP payloads of the TCP service packets in the traffic between the first TCP connection established by the application client and the third TCP service packet sent by the application server, RATMD achieves a detection accuracy of 95.5%.

Machine learning-based approaches for ubiquitination site prediction in human proteins

Protein ubiquitination is a critical post-translational modification (PTMs) involved in numerous cellular processes. Identifying ubiquitination sites (Ubi-sites) on proteins offers valuable insights into their function and regulatory mechanisms. Due to the cost- and time-consuming nature of traditional approaches for Ubi-site detection, there has been a growing interest in leveraging artificial intelligence for computer-aided Ubi-site prediction. In this study, we collected experimentally verified Ubi-sites of human proteins from the dbPTM database, then conducted comprehensive state-of-the art computational methods along with standard evaluation metrics and a proper validation strategy for Ubi-site prediction. We presented the effectiveness of our framework by comparing ten machine learning (ML) based approaches in three different categories: feature-based conventional ML methods, end-to-end sequence-based deep learning (DL) techniques, and hybrid feature-based DL models. Our results revealed that DL approaches outperformed the classical ML methods, achieving a 0.902 F1-score, 0.8198 accuracy, 0.8786 precision, and 0.9147 recall as the best performance for a DL model using both raw amino acid sequences and hand-crafted features. Interestingly, our experimental results disclosed that the performance of DL methods had a positive correlation with the length of amino acid fragments, suggesting that utilizing the entire sequence can lead to more accurate predictions in future research endeavors. Additionally, we developed a meticulously curated benchmark for Ubi-site prediction in human proteins. This benchmark serves as a valuable resource for future studies, enabling fair and accurate comparisons between different methods. Overall, our work highlights the potential of ML, particularly DL techniques, in predicting Ubi-sites and furthering our knowledge of protein regulation through ubiquitination in cells.