International Journal of Intelligent Systems最新文献

Application-layer distributed denial of service (DDoS) attacks have become the main threat to Web server security. Because application-layer DDoS attacks have strong concealability and high authenticity, intrusion detection technologies that rely solely on judging client authenticity cannot accurately detect such attacks. In addition, application-layer DDoS attacks are periodic and repetitive, and attack targets suddenly in a short period. In this study, we propose an efficient application-layer DDoS detection system based on improved random forest. Firstly, the Web logs are preprocessed to extract the user session characteristics. Subsequently, we propose a Session Identification based on Separation and Aggregation (SISA) method to accurately capture user sessions. Lastly, we propose an improved random forest classification algorithm based on feature weighting to address the issue of an increasing number of features leading to prolonged calculation times in the random forest algorithm, and as the feature dimension increases, there might be instances where no subfeature is related to the category to be classified. More importantly, we compare the request source IP with the malicious IP in the threat intelligence library to deal with the periodicity and repetition of application-layer DDoS attacks. We conducted a comprehensive experiment on the publicly available Web log dataset and the threat intelligence database of the laboratory as well as the simulated generated attack log dataset in the laboratory environment. The experimental results show that the proposed detection system can control the false alarm rate and false alarm rate within a reasonable range, improving the detection efficiency further, the detection rate is 99.85%. In secondary attack detection experiments, our proposed detection method achieves a higher detection rate in a shorter time.

The proliferation of photorealistic images synthesized by generative adversarial networks (GANs) has posed serious threats to society. Therefore a new challenge task, named image attribution, is arising to attribute fake images to a specific GAN. However, existing approaches focus on model-specific features but neglect the misguidance of semantic-relevant features in image attribution, which leads to a significant performance decrease in cross-dataset evaluation. To tackle the above problem, we propose a semantic-agnostic fake image attribution (SFIA) method, which effectively distinguishes fake images by disentangling the GANs fingerprint and semantic-relevant features in latent space. Specifically, we design a semantic eliminator based on residual block with skip connections that take images as input and outputs GAN fingerprint features. A classifier with an attention module for feature refinement is introduced to make the final decision. In addition, we develop a well-trained reconstructor and classifier which supervise the semantic eliminator to achieve semantic-agnostic feature extraction. Moreover, we propose an improved data augmentation combined with meta-learning to enhance the model’s generalization in detecting unseen image categories. Comprehensive experiments on various datasets, namely, CelebA, LSUN-church, and LSUN-bedroom, demonstrate the effectiveness of our proposed SFIA. It achieves over 95% accuracy on three datasets and exhibits superior performance in terms of generalization to unseen data.

Anonymous network tracing is a significant research subject in the field of network security, and flow correlation technology serves as a fundamental technique for deanonymizing network traffic. Existing flow correlation techniques are considered ineffective and unreliable when applied on a large scale because they exhibit high false-positive rates or require impractically long periods of traffic observation to achieve reliable correlations. To address this issue, this paper proposed an innovative flow correlation approach for the typical and most widely used Tor anonymous network by combining graph convolutional neural networks with triplet networks. Our proposed method involves extracting features such as packet intervals, packet lengths, and directions from Tor network traffic and encoding each flow into a graph representation. The integration of triplet networks enhances the internode relationships, which can effectively fuse flow representations with node associations. The graph convolutional neural network extracts features from the input graph topology, mapping them to distinct representations in the embedding space, thus effectively distinguishing different Tor flows. Experimental results demonstrate that with a false-positive rate as low as 0.1%, the correlation accuracy reaches 86.4%, showcasing a 5.1% accuracy improvement compared to the existing state-of-the-art methods.

Through modeling the characteristics of wireless transmission channels, channel estimation can improve signal detection and demodulation techniques, enhance the spectrum utilization, optimize communication performance, and enhance the quality, reliability, and efficiency of intelligent wireless communication systems. In this paper, we propose a deep convolutional autoencoder–based channel estimation method in intelligent wireless communication systems. At first, the channel time-frequency response matrix between the transmitter and receiver can be represented as 2D images. Then they are fed into the convolutional autoencoder to learn key channel features. To reduce the structural complexity of the deep learning model and improve its inference efficiency, we adopt the method of removing redundant parameters to achieve model compression. Iterative training and pruning based on stochastic gradient descent (SGD) and weight importance evaluation are alternated to obtain a lightweight deep learning model for channel estimation. Finally, extensive simulation results have verified the effectiveness and superiority of the proposed method.

Early warnings of enterprise credit risk based on supply chain scenarios are helpful for preventing enterprise credit deterioration and resolving systemic risk. Enterprise credit risk data in the supply chain are characterized by higher-dimension information and class imbalance. The class imbalance influences the feature selection effect, and the feature subset is closely related to the predictive performance of subsequent learning algorithms. Therefore, ensuring the adaptivity of feature selection and the subsequent class imbalance–oriented classification model is a key issue. We propose an AdaBoost ensemble model with fast nondominated feature selection (AdaFNDFS). AdaFNDFS uses the FNDFS method in the AdaBoost algorithm to iteratively select features and uses the classifier to evaluate the performance of feature subsets to train the class imbalance–oriented classifier and the best-matched feature subset, ensuring the adaptivity of feature selection and subsequent classifiers. The further use of the differential sampling rate (DSR) method enables AdaFNDFS to integrate more training models with different knowledge and to obtain higher accuracy and better generalization ability for prediction tasks facing high-dimensional information and class imbalance. A test using credit risk data from Chinese listed enterprises containing supply chain information demonstrates that the prediction scoring indicators, such as AUC, KS, AP, and accuracy, of the AdaFNDFS are better than those of basic models such as LR, LDA, DT, and SVM and multiple hybrid models that use SMOTE, feature selection, and ensemble methods. AdaFNDFS outperforms the basic models by at least 0.0073 (0.0344, 0.0349, and 0.0071) in terms of the AUC (KS, AP, and accuracy). AdaFNDFS has outstanding advantages in predicting enterprise credit risk in the supply chain and can support interested decision-makers.

The National Disaster Management Center has declared a drought disaster in the Northern Cape, South Africa, due to persistent dry conditions that impact regions such as the Western, Eastern, and Northern Cape provinces. Accurate drought predictions are vital for decision-making and planning in vulnerable areas. This study introduces a hybrid intelligence model, combining long short-term memory (LSTM) and convolutional neural networks (CNNs), to forecast short-term meteorological droughts using the Standardized Precipitation Evapotranspiration Index (SPEI). Applied to Kimberley and Upington in the Northern Cape, the model predicts 1-month and 3-month SPEI indices (SPEI-1 and SPEI-3). The hybrid model’s performance, compared to benchmark models such as artificial neural networks (ANNs), LSTM, and CNN, is measured through statistical analysis. In Kimberley, the CNN–LSTM model displayed a robust positive correlation of 0.901573 and a low mean absolute error (MAE) of 0.082513. Similarly, in Upington, the CNN–LSTM model exhibited strong performance, achieving a correlation coefficient of 0.894805 and a MAE of 0.085212. These results highlight the model’s remarkable precision and effectiveness in predicting drought conditions in both regions, underscoring its superiority over other forecasting techniques. SPEI, incorporating potential evapotranspiration and rainfall, is superior for drought analysis amidst climate change. The findings enhance understanding of drought patterns and aid mitigation efforts. The CNN–LSTM hybrid model demonstrated noteworthy results, outperforming ANN, CNN, and LSTM, emphasizing its potential for precise meteorological drought predictions.

In the present era, where competition pervades across all domains, profitability holds crucial economic importance for numerous companies, including the banking industry. Offering the right products to customers is a fundamental problem that directly affects banks’ net revenue. Machine learning (ML) approaches can address this issue using customer behavior analysis from historical customer data. This study addresses the issue by processing customer transactions using a bank’s current account debt (CAD) product with state-of-the-art ML approaches. In the first step, exploratory data analysis (EDA) is performed to examine the data and detect patterns and anomalies. Then, different regression methods (tree-based methods) are tested to analyze the model’s performance. The obtained results show that the light gradient boosting machine (LGBM) algorithm outperforms other methods with an 84% accuracy rate in the light gradient boosting algorithm, which is the most accurate of the three methods used.

With the increasing utilization of cameras and three-dimensional Light Detection and Ranging (LiDAR) systems in perception tasks, the fusion of these two sensor modalities has emerged as a prominent research focus in the fields of robotics and unmanned systems. While various extrinsic calibration methods have been developed, they often suffer from limited accuracy when using low-resolution LiDAR sensors and require the placement of calibration targets at multiple locations. This paper introduces a novel calibration target known as the Three-Dimensional Towered Checkerboard (3TC), along with a precise and straightforward extrinsic calibration approach for camera-LiDAR systems. The 3TC consists of stacked cubes adorned with planar or 2D checkerboards, which provide the known positions of checkerboard corner points in three-dimensional space. Leveraging the Iterative Closest Point (ICP) algorithm, the proposed method calculates the spatial relationship between LiDAR point cloud data and the 3TC model to infer the positions of checkerboard corner points in the LiDAR coordinate system. Subsequently, the Perspective-n-Point (PnP) algorithm is employed to establish the correlation between corner positions in the LiDAR coordinate system and the camera image, given the intrinsic parameters of the camera. By ensuring an adequate number of cubes and 2D checkerboards on a specific 3TC, along with accurately estimated corner point positions in LiDAR, a single frame of data from both the camera and LiDAR facilitates their extrinsic calibration. Experimental validations conducted across diverse camera and LiDAR systems, achieving minimal error close to the theoretical limit of the devices, attest to the robustness and precision of the 3TC and the proposed calibration methodology.

Federated analytics (FA) over spatial data with local differential privacy (LDP) has attracted considerable research attention recently. Existing solutions for this problem mostly employ a uniform grid (UG) structure, which recursively decomposes the whole spatial domain into fine-grained regions in the distributed setting. In each round, the sampled clients perturb their locations using a random response mechanism with a fixed probability. This approach, however, cannot encode the client’s location effectively and will lead to ill-suited query results. To address the deficiency of existing solutions, we propose LDP-FSRQ, a spatial range query algorithm that relies on a hybrid spatial structure composed of the UG and quad-tree with nonuniform perturbation (NUP) probability to encode and perturb clients’ locations. In each iteration of LDP-FSRQ, each client adopts the quad-tree to encode his/her location into a binary string and uses four local perturbation mechanisms to protect the encoded string. Then, the collector prunes the quad-tree of the current round according to the clients’ reports and shares the pruned tree with the clients of the next round. We demonstrate the application of LDP-FSRQ on Beijing, Landmark, Check-in, and NYC datasets, and the experimental results show that our approach outperforms its competitors in terms of queries’ utility.

Nonlinear optimization is becoming more challenging in information sciences and various industrial applications, but nonlinear problems solved by the classical particle swarm-based methods are usually characterized by low efficiency, accuracy, and convergence speed in specific issues. To solve these problems and enhance the nonlinear optimization performance, an improved metaheuristic particle swarm optimization (PSO) model is proposed here. First, the optimization principles and model of the new method are introduced, and algorithms of the improved PSO are presented by updating the displacement and velocity of the moving particle according to Euler–Maruyama (EM) principle rather than traditional standard normal distribution. Then, the influence of the model parameters, input dimensions, and different nonlinear problems on the PSO optimization characterizations are studied by Pareto set solving and optimization performance comparison. The analysis regarding diverse nonlinear problems and optimization methods manifests that the improved method is capable of solving various nonlinear problems especially for multiobjective models, while the robustness and reliability can always keep consistent regardless of the change of model parameters. Finally, the performance evaluation is exhibited by the case study of nonlinear parameter optimization, 3 groups of CEC benchmark problems, and rank-sum test for 6 comparable optimization algorithms, which all verify its effectiveness and reliability, as well as the significance and great application promise. The results show that the new proposed PSO method has the fastest convergence speed and least iteration numbers in searching for the global best solution of 9 nonlinear problems among 8 different optimization models indicated by the p values smaller than 0.05. Additionally, the main conclusions showing the calculation efficiency, stability, robustness, and great application promise of the proposed method are summarized, and future work is discussed.