This paper discusses our plan for an education program for data driven security for machine learning applications. We discuss the motivation for such a program, details of the program, various course modules to be developed as well as modules to be incorporated into our current courses in cyber security.
Recent digital transformation across industries has extended the demands for secure cross-domain transactions and data sharing. However, traditional Public Key Infrastructure (PKI)-based systems encounter challenges deriving from implementing centralized identity management setting, such as single-point failures, data silos, and insufficient privacy safeguards. Blockchain-based Decentralized Identity (DID) schemes have emerged as a promising solution to addressing cross-domain authentication challenges. In this work, we have systematically examined DID applications in cross-domain authentication, analyzing core mechanisms, e.g., privacy-preserving techniques, anonymous credentials, and lightweight verification protocols. We also investigate a few key supportive technologies, including Zero-Knowledge Proofs (ZKP) and cryptographic accumulators. Main findings of this work covers identifying consensus mechanism limitations and suggesting future potential research directions.
The welfare, productivity, and health of livestock are being impacted steadily by climate change and extreme weather events. For instance, it is predicted that heat stress will cause dairy cows in Southeast England to lose more than 170 kg of milk a year; in the hottest UK regions, this loss could increase to 1,300 kg per cow (18.6% of annual yield) by the 2090s. A solid foundation for evaluating environmental stress is presented in this paper. It is based on agro-environmental indices that are obtained from temperature and humidity data. To improve predictive capabilities, the proposed Virtualized Farming Analytics for Smart Decision-Making (V-Farm) system uses state-of-the-art synthetic data generation techniques, such as TCGAN and polynomial fitting. A novel approach to lessen the effects of thermal stress in precision livestock farming is to incorporate these indices into automated ventilation systems.
Blockchain technology, providing security, trust, and transparency for a variety of applications, has become a revolutionary force in decentralized systems. However, scalability continues to be a major obstacle, restricting its use in large-scale and high-throughput settings. To assess the trade-offs between decentralization, security, and computing efficiency, our survey provides a thorough examination of both on-chain and off-chain scalability options, such as sharding, adaptive block sizing, rollups, state channels, and sidechains. Layered protocols (Layer 1 and Layer 2) are also discussed in terms of scalability and interoperability. By combining on-chain, off-chain, and modular blockchain designs, this work investigates new hybrid models, multi-layer architectures, and cross-chain scalability mechanisms in addition to conventional solutions. By identifying critical gaps and challenges in existing approaches, this survey provides a roadmap for future research in scalable, secure, and interoperable blockchain infrastructures, with applications in decentralized finance, IoT, and cross-border transactions.
The significance of privacy and security in online social networks is to protect sensitive information, prevent unauthorized access, and build trust among users while ensuring a secure online environment. This paper proposes a friendship matching system based on the Parlier encryption algorithm, aiming at addressing existing privacy breaches and data misuse issues on current friendship platforms, significantly improving matching accuracy. First, filter the data items that users care about more in the database so as to make the appropriate friends. Then, according to the homomorphic encryption of paillier algorithm, in the case of ciphertext, the weight of each data item is calculated to the selected users. Finally, the Euclidean distance is used to calculate the similarity between users and select the appropriate dating users. Experimental results demonstrate that the proposed system exhibits strong performance and feasibility in preserving user privacy and data security.
The demand for more bandwidth capacity is growing in response to the rise in mobile users and bandwidth-intensive services, which makes massive MIMO systems essential. This leads to a need to develop these systems further using deep learning to address the complexity of wireless channels and dynamically optimise resource allocation. However there are lots of initiatives that utilise deep learning approaches in 5G signal detection for massive MIMO, there is a research gap in comprehensive comparative analyses across multiple signal detection techniques, various MIMO sizes, and a wide range of SNR levels to provide a holistic understanding of their performance in real-world scenarios. This paper proposes a deep learning approach for enhancing the massive MIMO. A deep neural network is utilised for detecting the symbols in the massive MIMO setting, called deep MIMO. This system facilitates sequential symbol detection through an unfolded network architecture, enabling the processing of data in a layer-by-layer manner. With intensive training and optimisation, the model achieves better performance in massive MIMO scenarios.
Network covert channels facilitate the transmission of secret information in an undetectable manner. In this survey, we systematically study fundamental concepts, construction techniques, and detection methods of network covert channels. The channels are categorized into two types, namely, storage-based covert channels and timing-based covert channels. The former utilizes redundancy in protocol fields to embed information, while the latter transmits data by modulating traffic intervals. The timing-based channel primarily enhances concealment by adjusting the packet intervals and integrating protocol features, while the storage-based channel embeds information using redundant fields of each protocol layer. In terms of detection methods, timing-based channels are identified through statistical analysis of traffic patterns combined with machine learning models. Conversely, storage-based channels are typically uncovered by implementing protocol compliance verification and deep packet inspection of header fields and payload data. This survey investigates the evolution path of network covert channel construction and detection technologies in details, providing a theoretical foundation for the design of future network covert communication solutions.
With the development of power grid operations, the need for advanced natural language processing (NLP) techniques to analyze and manage vast amounts of data has become increasingly critical. However, the sparsity and complexity of power grid operations data pose challenges for general large- scale language models (LLMs) to effectively perform analysis tasks. Additionally, the general LLMs, which lack of professional knowledge in the power grid operations domain, produce highly irregular and ineffective solutions. To address these challenges, this paper proposes an efficient fine-tuning method PGrid-Align, which contains two stages. In stage one, the pre-processing method based on prompt engineering is introduced to handle sparse professional data. This method uses prompts with classification identifiers to summarize and organize sparse power grid operations data, helping general LLMs understand the special data structure and complete analysis tasks. In stage two, an efficient fine-tuning method for power grid data is introduced. This method references the classification identifiers from stage one and performs knowledge-level QLoRA fine-tuning on the LLM, improving its logical alignment. To demonstrate the effectiveness of PGrid-Align, this paper designs analysis tasks using sparse power grid operations data. The experiment results show that LLMs fine-tuned with PGrid-Align can complete the analysis tasks in a standardized manner and high accuracy, reflecting its effectiveness and potential for real-world power grid applications.
Blockchain covert channels have become a cutting-edge technique of information concealment that allows anonymous communication, censorship resistance, and privacy protection. The processes of blockchain-based covert communication are methodically examined in this survey, such as information embedding, transaction screening, and transaction obfuscation. In particular, this work analyzes temporal and storage covert channels in blockchain covert communication embedding mechanisms. To improve security and thwart discovery, transaction screening methods are also investigated, covering dynamic label creation, encryption-based screening, and fixed address screening. In order to enhance concealment, we have studied obfuscation strategies at several levels, including transaction fields, structures, and patterns. Blockchain covert communication has advanced, but there are still a few issues, which involve computational complexity, trade-offs between efficiency and secrecy, and a limited embedding capacity. This work offers a thorough analysis of blockchain covert channels and explores the means of adoptability and efficacy.
In the era of big data, ontology matching, which determines correspondences between concepts, has become a crucial knowledge-mining task. Existing ontology matching methods typically focus on semantics, structure, and instances, often optimizing datasets within specific domains. However, these methods suffer from limited universality when applied across diverse domains. To address this issue, this paper proposes a Robust Ontology Matching method based on Ensemble Learning (ROME). By integrating multiple machine learning algorithms, ROME improves scalability, robustness, and performance, overcoming the challenges of noisy data and data distribution shifts. The proposed approach employs ensemble learning techniques, such as Bagging and Boosting, to enhance model generalization and achieve superior results in ontology matching. Extensive experiments on human and rat anatomy datasets demonstrate that the ensemble fusion approach outperforms existing methods, yielding better mapping precision, recall, and F1 scores.
With the arrival of the digital age and the wide adoption of Internet technologies, the manufacturing industry is gradually entering the era of smart manufacturing. It is of great significance to adopt efficient digital solutions for data management. Organizations tend to choose distributed third-party cloud platforms to store and share data to reduce the costs of building their data platforms. Although distributed cloud storage solutions can solve various problems associated with centralized storage solutions, the collaboration between multiple nodes requires mutual trust. Blockchain technology can help to build trust among different parties. However, the existing consortium blockchain platform does not meet the needs of data privacy and large-scale data storage in smart manufacturing. Therefore, we propose a storage scheme based on the consortium blockchain platform Hyperledger Fabric and the InterPlanetary File System (IPFS). By improving the ledger data structure and designing corresponding smart contract functions, the proposed scheme can store various digital resources on the blockchain platform flexibly and provide fine-grained protection for sensitive data. Several simulation experiments are conducted to analyze the factors affecting the performance. The experiment results show that the space-time overhead of the proposed scheme is acceptable.
With the deep penetration of cloud computing in the financial industry, the convergence amount of financial data in the cloud continues to climb, and the problem of data security has become a bottleneck restricting the development of the financial data sharing. Traditional security protection means, such as access control, data encryption and storage, etc., gradually expose their limitations in the face of increasingly complex network attacks and diversified data use scenarios, and homomorphic encryption technology arises at the historic moment. This paper constructs a financial data security sharing scheme including financial institutions, financial certification center, Credit Data Statistics Center and financial key management center. The scheme first encrypts customer credit data and uploads it, then computes the encrypted data, and finally returns the ciphertext result to be decrypted by financial institutions. It not only ensures data privacy, but also realizes data sharing, and provides practical methods for financial institutions to prevent non-performing loans. As standardization efforts progress and computational overheads diminish, homomorphic encryption is expected to underpin next-generation secure financial infrastructures, fostering innovation in personalized financial services and regulatory compliance frameworks.
Fire is a disaster with great harm in real life, and traditional fire detection methods have many shortcomings. In recent years, the development of computer vision and deep learning has provided new solutions for fire detection. This paper proposes a fire monitoring method based on the YOLOv5 target detection algorithm, which uses monitoring equipment to achieve real-time fire detection. The research content includes the construction of self-built data sets, YOLOv5 model training and optimization, detection system design and performance verification. By embedding the CBAM module in the C3 layer of the YOLOv5s backbone network, the channel and spatial processing of the feature map is performed, the feature refinement is achieved, and the construction and visualization of the detection system are completed. The experimental results show that the detection accuracy of flame targets is 0.745, and the detection accuracy of smoke targets is 0.582; on the test set, the average detection accuracy of flames and smoke is 0.823 and 0.673 respectively, and the overall average detection rate is 0.748. The fire detection system based on YOLOv5 implemented in this work has a significant effect on the current fire prevention and control, and it meets the actual application scenarios to the greatest extent.
Cloud computing dramatically enhances digital resource delivery as workload patterns keep changing and hinder effective resource sharing. This research develops an efficient resource management solution using machine learning algorithms, including a random forest classifier, a gradient boosting classifier, and a support vector classifier. This study analyses CPU usage, memory usage, and network traffic in the Cloud Computing Performance Metrics dataset to create precise demand forecasts. Our model results improved after applying data preparation steps to normalize and engineer the features. The new system dynamically allocates resources based on predictive models, helping the platform reduce cost and improve delivery speed. Our test results show that asset management performs better than fixed resource distribution and load-spreading systems, especially in demanding cloud environments.
Driven by global carbon neutrality goals, carbon data sharing in power systems has become a critical component of low-carbon transformation. However, the high sensitivity and privacy requirements of carbon data pose significant challenges to traditional data security technologies. To address these challenges, this paper proposes an innovative lattice-based attribute encryption scheme designed to enhance the security and privacy of carbon data sharing in power systems. The scheme integrates the quantum-resistant properties of lattice-based cryptography with the fine-grained access control mechanisms of attribute-based encryption, enabling efficient encryption and secure sharing of carbon data. Through theoretical analysis, this paper demonstrates the scheme's superiority in resisting quantum computing attacks and supporting dynamic access policies. Experimental results indicate that the proposed scheme outperforms existing methods in terms of encryption efficiency (approximately 30% improvement), decryption success rate (over 99.5%), and resistance to attacks (withstanding 99.9% of simulated attacks). Furthermore, in large-scale power system environments, the scheme exhibits excellent scalability, supporting concurrent data sharing for over 10,000 nodes with a system latency of less than 200 milliseconds. This research not only provides an efficient and secure solution for carbon data sharing in power systems but also offers new insights into the application of lattice-based cryptography in real-world industrial scenarios, with significant theoretical and practical implications.
As the number of unmanned aerial vehicle (UAV) operations is growing rapidly, the risk of collisions increases significantly, making the coordination and verification of flight path compliance crucial. Since many different stakeholders, such as different UAV service suppliers (USS) and UAV operators are involved in an advanced air mobility (AAM) system, the system shall be decentralized and telemetry data shall be measured by the local community using sensor devices. In order to increase system resilience, sub-components of the system are implemented on a blockchain. Smart contracts are used to check whether a UAV has actually navigated the route specified by the USS pre-flight. Due to the restrictions of the system, the flight plan cannot be publicly revealed. Zero-Knowledge proofs (ZKPs) are unfeasible for this use case due to the high number of transactions and computational effort. Therefore, a novel approach has been developed that crosschecks measured telemetry data of UAV flights with flight plans and verifies the correctness without revealing any sensitive flight information. The verification results of UAV telemetry data can further be used to reward UAV operators for complying with the planned flight path.
Attacks that aim to disrupt services, known as Distributed Denial of Service (DDoS), exploit protocol vulnerabilities and utilize legitimate traffic to target systems, thereby threatening the normal operation of the network. This paper proposes a method for filtering DDoS attack traffic along routing paths within the Software Defined Networking (SDN) domain. The method, based on the SDN framework, employs a consistency hashing algorithm to select nodes in the global routing platform, forming secure access paths. By utilizing the accuracy of edge node detection and inter-domain cooperation, the attack is effectively mitigated along the path close to its origin, which is referred to as path filtering. The paper also provides a design for the filtering framework and details the filtering steps.
Intelligent reflecting surfaces (IRS) have been extensively employed to enhance physical layer security. However, passive IRS face limitations in enhancing secrecy performance due to the multiplicative attenuation effect. To address this limitation, we propose an innovative design leveraging active IRS to enhance the secure computing performance of the mobile edge computing (MEC) system in the presence of eavesdropper. Furthermore, wireless power transfer (WPT) technology is integrated to furnish a continuous energy supply to edge devices, thereby facilitating the execution of computational tasks. In the proposed framework, energy signals are initially transmitted by the base station (BS) to charge multiple users. The harvested energy is subsequently used by the users for local computation and partial task offloading in the presence of eavesdropper. Simultaneously, active IRS is deployed to enhance energy harvesting efficiency and secure task offloading. To maximize the secure computation rate, we jointly optimize the time allocation for WPT and task offloading, the energy and receive beamforming of the BS, the reflection coefficients of the active IRS, the user’s local computation frequency, and the transmit power. The resulting non-convex optimization problem is addressed through an iterative algorithm based on block coordinate descent (BCD). Simulation results indicate that the proposed active IRS-based scheme achieves substantial improvements in secure computation rate compared to existing benchmark methods.
Efficient collaborative operation of high orbit remote sensing satellite data transmission and processing equipment is an effective guarantee for the efficient production, quality and cost of remote sensing data image products. In order to improve the operational efficiency of remote sensing data transmission and processing system equipment, optimize equipment configuration schemes, save equipment cost and accelerate data processing efficiency, this paper builds a network transmission simulation model based on virtualization technology in response to the high real-time and large data volume characteristics of high orbit satellite observation. Based on Arena software, simulation research was conducted on the operational efficiency of system processing equipment. Combined with the specific engineering example of Gaofen-4 satellite, model parameter setting and simulation are carried out, and the operational efficiency under different equipment configuration schemes are analyzed. The study provides a reference for decision-making of the actual operation of high orbit remote sensing satellite data transmission and processing.
In the face of rapidly evolving network attacks and the large volume of network traffic data, distributed intrusion detection has garnered significant research interest. However, distributed intrusion detection requires multiple parties to share data, which may pose a challenge of sensitive data leakage. This paper presents IDS-BF, a distributed intrusion detection system based on blockchain and federated learning. To improve the effectiveness of federated learning, the proposed system adopts a contribution-based method to dynamically adjust the weights of clients when aggregating the local gradients. Specially, to ensure the fairness of contribution evaluation, the proposed system utilizes two parachains to perform model aggregation and contribution evaluation respectively. With the help of a relay chain, the two parachains can communicate with each other. Simulation results on real-world data show that the proposed aggregation strategy can help to improve the accuracy of global model, achieving an increase of up to 8% compared to traditional strategies.
