Protection against Webshell attacks is an important issue in the field of network security. Attackers often design various anti-virus techniques when writing Webshell to bypass firewalls and evade detection. Therefore, existing methods have problems with low accuracy and high false positives in Webshell detection. To address these issues, this paper proposes a Webshell detection method based on CodeBERT and deep learning. Firstly, the sample is analyzed and preprocessed by Antlr4 syntax analysis. Then, the preprocessed sample is encoded by CodeBERT model, and the sample is converted into feature vector. Finally, the BiGRU-Attention model is used for training and classification, so as to complete the detection of the sample. The method used in this paper is compared with the mainstream method. The experimental results show that the method proposed in this paper has better effect for Webshell detection and lower false positive rate.
This study aimed to determine the extent to which Jordanian journalists in digital newspapers are aware about artificial intelligence (AI) techniques in journalistic practices and to identify the most important applications, difficulties, and motives for employing AI from the perspectives of 59 journalists in Jordanian digital newspapers. An online questionnaire was used. The results were as follows: the weak reality of employing AI techniques in journalistic practices; few journalists know about the technologies sufficiently; and the most used systems in journalistic practices are photo and video verification techniques, followed by breaking news tracking techniques. The findings revealed that economic factors and the absence of technological infrastructure in press institutions are among the most important factors affecting the adoption of AI techniques in Jordanian digital newspapers. This study had some limitations that should be addressed in future studies.
This paper aims to reflect on the extent to which the Design Science Research Method (DSRM) is aligned with orchestrating the development of cancer care data analytics research projects using a data strategy-enabled use-case-driven approach. A fit-for-purpose DSRM research framework has been designed to guide the research design and project-manage the iCanViz research project. The iCanViz project is an intelligent cancer data analytics visualizer aligned to multiple cancer sites, cancer incidences, risk correlations, survivals, and deaths in the Arab world within certain years of reported cancer incidences per cancer site. The DSRM approach has been deemed closely aligned with the iCanViz research project. This is primarily due to its ability to articulate and incorporate cancer care data strategy use-cases as agile software development increments within the iCanViz DSRM research framework. Also, the adaptation of the iCanViz DSRM research framework has contributed to enhancing the even distribution and concurrent execution of cancer care data analytics tasks across the software development and research team. Furthermore, the iCanViz DSRM research design has proven effective in designing research frameworks that bridge the gap between the world of cancer care and the creation of cancer care data analytics systems. With a small research team, the iCanViz project achieved this within a relatively short time frame of less than a year.
This paper presents a nonlinear impairment compensation method for an integrated sensing and communication system (ISAC) in optical fibers. The proposed method is based on learned digital back-propagation algorithm (LDBP). The optical domain ISAC waveform used in this work, i.e., 4-PAM-LFM, is generated by modulating a linear frequency modulation (LFM) optical carrier with a 4-level pulse amplitude-modulation signal. The LFM optical carrier makes the dispersion and nonlinear impairment characteristics of 4-PAM-LFM very different from those of the conventional optical communication waveforms (e.g., 16-QAM). The effectiveness of LDBP for 4-PAM-LFM is verified by building a numerical simulation platform. The influence of initial frequency, bandwidth, and launch power of 4-PAM-LFM on the compensation effectiveness of LDBP is also examined. It is observed that the LDBP compensation results become obviously worse when the initial frequency is increased. On the contrary, the bandwidth of 4-PAM-LFM only has a slight impact on the LDBP compensation results. This work provides a beneficial reference for the design of optical-domain ISAC waveforms.
This paper focuses on the classification of secure IEC 104 protocol traffic in energy systems using a specific convolutional neural network model. Secure communication of the IEC 104 protocol was used to train the network. The data were obtained using a special network traffic simulator and from an energy testbed. In order to analyze secure communication, a classifier was developed to identify the individual operating states of the communicating station. In this article, we focused on the classification of IEC 104 protocol communication with TLS security. The classifier consisted of a convolutional neural network with a defined two-dimensional input matrix. The matrix was composed of the information from five consecutive packets. The information was constructed from the interarrival time between packets, the length of TLS encrypted application data, and the encrypted application data up to 64 B in size. To obtain enough data to train the convolutional network, a simulator of characteristic messages for each state was developed. The classifier was trained to accurately classify the "Normal operation" and "Short circuit" states of the station, achieving a probability exceeding 90 % for the distinct data flow. However, in the case of other operating states characterized by subtle differences, misclassification occurred between two states sharing similar characteristics.
The widespread use of UAVs in smart cities for tasks like traffic monitoring and environmental data collection creates significant privacy and security concerns due to the transmission of sensitive data. Traditional UAV-MEC systems with centralized data processing expose this data to risks like breaches and manipulation, potentially hindering the adoption of these valuable technologies. To address this critical challenge, we propose UBFL, a novel privacy-preserving federated learning mechanism that integrates blockchain technology for secure and efficient data sharing. Unlike traditional methods relying on differential privacy (DP), UBFL employs an adaptive nonlinear encryption function to safeguard the privacy of UAV model updates while maintaining data integrity and accuracy. This innovative approach enables rapid convergence, allowing the base station to efficiently identify and filter out severely compromised UAVs attempting to inject malicious data. Additionally, UBFL incorporates the Random Cut Forest (RCF) anomaly detection algorithm to actively identify and mitigate poisoning data attacks. Extensive comparative experiments on benchmark datasets CIFAR10 and Mnist demonstrably showcase UBFL’s effectiveness. Compared to DP-based methods, UBFL achieves accuracy (99.98%), precision (99.93%), recall (99.92%), and F-Score (99.92%) in privacy preservation while maintaining superior accuracy. Notably, under data pollution scenarios with varying attack sample rates (10%, 20%, and 30%), UBFL exhibits exceptional resilience, highlighting its robust capabilities in securing UAV gradients within MEC environments.
Authentication is the first line of defense in communication security. Physical layer authentication (PLA) based on Channel Impulse Response (CIR) offers a lightweight approach. Unlike traditional cryptographic-based authentication, physical layer authentication identifies data packets through wireless CIRs. This avoids the compatibility requirements of upper-layer protocols and presents advantages such as flexibility in authentication methods and suitability for massive heterogeneous terminal access. However, current physical layer authentication often requires illegal CIR samples, which are hard to obtain. Furthermore, training for deep learning-based physical layer authentication requires numerous samples, and the low latency characteristics of Mobile Edge Computing (MEC) means it might not have ample time to gather sufficient signals, reducing authentication performance. This paper presents a physical layer authentication scheme based on deep learning, mapping CIRs to device locations, and further mapping to their authentication, achieving multi-user physical layer authentication. In the absence of illegal device CIR samples, it discriminates against illegal devices and classifies legal devices. Moreover, a data augmentation technique using noise injection is employed to enhance the performance of the multi-user physical layer authentication scheme, speeding up the training process. Extensive experiments on public datasets confirm the efficacy of our approach.
The Flying Ad-hoc Networks (FANETs) is interconnected by multiple small drones in a multi-hop self-organizing way, which can perform some tasks in special environments, due to the open wiress boundary , drones are vulnerable to adversary attacks. But how to protect the identity of the drone from being leaked to achieve secure communication between the drone and the ground station is a major challenge. To effectively address this issue, a certificateless aggregated signature scheme for Flying ad-hoc network is proposed. A drone with high computer aggregate signed messages from regular drones and upload them to a Trusted Authority (TA), verified by the TA and sent to the ground station, TA generates pseudo-identities for drone users to achieve anonymity. The use of aggregated signature technology reduces computational and communication overhead. Finally, based on the Hyperelliptic Curve Cryptography (HECC) Discrete logarithm problem it is proved that the scheme is existentially unforgeable under adversary adaptive message selection attacks.
In networks with critical industrial processes where operational integrity is paramount, device identification is crucial for security and effective management. Without such identification, the potential for mismanagement and security breaches increases. Active scanning for network device identification poses risks, especially in industrial settings. Such scanning can disrupt operations or even cause damage. Therefore, finding non-invasive identification methods that bypass active scanning is imperative. Passive scanning, owing to its non-intrusive approach, is favored for industrial devices. Modern statistical learning techniques combined with passive scanning can mitigate risks of active methods. Our research harnesses time delay data in network communications to accurately identify specific industrial PLC models. We derive our data from timestamp details of the OPC UA protocol, widely recognized as a standard in industrial communication. Statistical variables from time delay data enhance the accuracy of passive device identification in industrial settings.
This paper explores the problem of optoelectronic task processing based on multilevel collaborative computing networks. The core keywords of the research include optoelectronic task processing, edge computing, cloud computing, and computational offloading. The aim of the article is to investigate how to efficiently process optoelectronic tasks in a multilevel collaborative computing network through an optimization approach. We propose a distribution-based algorithm to solve the minimum overhead of optoelectronic task processing. Further, we propose an entropy decision-based algorithm to improve existing algorithms to avoid falling into local optimal solutions.
With the rapid proliferation of wireless devices, effectively authenticating legitimate users has become a pivotal challenge in wireless communication. Amongst various approaches, physical layer authentication technology based on deep learning has garnered substantial attention from numerous researchers. In this paper, we propose a scheme for implementing physical layer authentication based on the Swin Transformer, utilizing Channel State Information (CSI) to distinguish between legitimate and illegitimate nodes in industrial network systems. In contrast to traditional physical layer authentication methods based on thresholds, the method proposed in this paper eschews the use of thresholds to achieve authentication. Moreover, compared to other methods based on deep neural networks, the introduction of attention mechanisms enables superior learning of wireless channel state features, enhancing model accuracy and reducing computational complexity. The efficacy of this scheme is validated through channel probing results in typical industrial wireless environments provided by the National Institute of Standards and Technology (NIST), which will facilitate the application of deep learning technology to industrial wireless network systems to enhance their security.
Aimed at the problem of APT attack-defense strategy selection in high-level network, through the analysis of APT attack process and stage, the network attack-defense game is constructed by using game theory, and a strategy selection method based on multistage APT attack-defense stochastic game (MA-ADSG) is proposed, which solves the problem the mixed strategy is not practical in the form of probability. The utility quantization calculation method is optimized, and the optimal defense strategy set is obtained by the strategy optimization algorithm. Finally, the feasibility and practicability of the method are verified by simulation experiment, which provides reference for high-level network defense.
The manufacturing and critical infrastructure world is increasingly dependent on cyber security. Cyber-attacks are becoming increasingly sophisticated, targeting traditional Information Technology (IT) and Operational Technology (OT) systems mainly due to the convergence of these infrastructures. Cyber-range assists with educating professionals and stakeholders about current trends and techniques by simulating realistic scenarios. This paper presents a scenario using the BUTCA (Brno University of Technology Cyber Arena) platform, developed at the Department of Telecommunications with a focus on IT and OT. The presented solution provides an effective and entertaining form of education using gamification aimed at a sector of interested students from primary school students to university students, professionals and interested members of the general public. As a proof of concept, we have developed a game (within BUTCA) targeting an industrial sector. Where the main objective of the cyber game is to demonstrate the potential impact of a successful cyber attack caused by human error through a phishing attack.
Because of the problem that the current industrial Internet of Things authentication scheme mainly uses the traditional identity authentication scheme, it is difficult to apply to the secure interaction of cross-domain terminals in the Industrial Internet of Things. This paper combines the trust characteristics of the trust domain and proposes an identity-based random challenge code authentication protocol based on lattice theory. It not only completes the identity authentication of the industrial Internet of Things terminals, but also realizes the trust authentication of the industrial Internet of Things terminals, ensuring the security of the industrial Internet of Things terminals security certification.
Addressing the inadequacies of existing trust models for communication data attacks among mobile network devices and their vulnerability to fraudulent activities from malicious device nodes, a dynamic trust assessment model is proposed for mobile network device nodes. In the computation of the direct trust model, consistency in data transmission among neighboring nodes is considered, thereby introducing reward and penalty factors to incentivize legitimate behavior of nodes. In the recommended trust model, the trustworthiness of a node is evaluated by measuring the dissimilarity between the source node and the recommended node. In the computation of the comprehensive trust model, both direct trust and recommended trust are integrated, and a dynamic weight allocation function is introduced to enhance the adaptability of the trust model.
The development of machine learning has brought new methods for botnet detection. Traditional machine learning methods and deep learning methods are used in botnet detection, but the former requires prior knowledge to select features. And deep learning methods solve this problem. This paper proposes a new botnet detection model that combines Convolutional Neural Network (CNN) with Support Vector Machine (SVM). This approach directly acquires network traffic data, preprocesses the data to obtain input suitable for CNN, utilizes CNN for feature extraction after preprocessing, and feeds the extracted features through two convolutional layers. The obtained features are then input to a SVM for classification. This method leverages the powerful feature extraction capabilities of CNN and the faster computational speed of linear SVM, resulting in faster training and excellent classification performance. Experiments show that the method has good performance on botnet detection and reduces training time compared to the CNN model.
This paper proposes an improved Sparrow Optimization Algorithm (ISOA) for optimizing the deployment of gateway nodes in wireless mesh networks (WMNs). The objective of this research is to minimize the number of gateways while maximizing the network throughput, minimizing network latency, and ensuring maximum coverage and connectivity for all nodes in the network. The proposed algorithm combines the foraging behavior and information exchange mechanisms of sparrows with the design of the objective function and constraints, resulting in an efficient and effective strategy for gateway node deployment. This paper provides a detailed explanation of the algorithm's improvement principles, implementation steps, performance evaluation, and experimental results. Comparative experiments and performance analysis are conducted to validate the superiority and effectiveness of the algorithm in addressing the optimization deployment problem of gateway nodes in wireless mesh networks.
The creation of suitable and efficient tools for anomaly detection constitutes a crucial aspect of security, applicable not only to industrial networks but also to cyber-physical systems. This article elucidates a framework designed to automate the selection of an optimal deep neural network architecture, thereby expediting the creation and implementation of neural network-based tools. The framework presented here enables a rapid design of an Artificial Neural Network structure without necessitating user intervention. Its efficacy has been showcased through experimentation with the publicly accessible HAI dataset, yielding an accuracy of approximately 0.94 after 10 epochs. Subsequently, a second scenario was performed where a total of 5456 models were generated and trained, with an average time of approximately 9.95 seconds per model.
Currently, intranet security detection faces several technical challenges: including an extremely imbalanced sample size and a small sample problem. We propose a model based on image and deep migration learning to address these problems. The single-day characteristics of users are modeled and transformed into grayscale images, which not only retain enough multivariate user behavior data, but also allow for intuitive analysis of the correlations between behaviors. To process user data, we use the traditional oversampling method, SMOTE (Synthetic Minority Oversampling Technique), and the SMOTE algorithm to ensure that the number of positive and negative samples is balanced. In order to increase the network's generalization capacity, a residual network pre-trained on the ImageNet dataset is applied to employee behavior detection We have performed a series of experiments with publicly available datasets to access the performance of the proposed scheme. The results show that this model performs exceptionally well in intranet security detection for resolving these problems.
This paper studies the problem of multi-unmanned aerial vejicles (multi-UAVs) cooperative search under communication constraints. We propose a distributed collaborative search method. In this method, we bulid target probability map for each UAV, and update the maps according to the Bayesian formula and the consensus algorithm. Based on the Shannon entropy and the target probability map, we design the objective function to guide UAVs fly to unexplored areas while keep the consistence of the communication network. Every UAV act as an independent agent, uses distributed model predictive control (DMPC) algorithm and particle swarm optimization (PSO) algorithm to achieve the trajectory optimazation. Finally, we made several simulation experiments to verify the feasibility of our method.
