ABSTRACT A zero‐knowledge proof lets one party convince another that a claim is true while withholding everything that would explain why it is true. We move that idea off conventional hardware and into chemistry, encoding a proof of graph isomorphism directly in synthetic DNA. Each node of a graph is given its own deliberately orthogonal DNA strand; an edge is confirmed only when a short complementary half‐linker meets its matching pair and forms a stable duplex. The verifier watches which bindings occur, but the pattern of binding never reveals how the two graphs line up, so the isomorphism stays hidden. Whether such a construction stays secure at the molecular level turns on two things: how distinguishable the sequences are, and how stable the duplexes they form turn out to be. We probe both. A seeded Monte Carlo study of orthogonal 20‐m libraries, built with balanced GC content and a minimum Hamming separation of , places the chance that an off‐target strand passes for a genuine linker on the order of : empirically at a binding threshold of mismatches, and under once the threshold is tightened to , each value reported with a Wilson confidence interval. This molecular error never becomes the bottleneck. A cheating prover already passes a round with probability one‐half from the isomorphism challenge alone, so the biochemical term enters soundness only as an additive correction, over the edges examined, rather than racing the decay across rounds. Read this way, DNA strands behave as cryptographic witnesses whose noise is small enough to bound and to account for, which lets a proof run at molecular scale without surrendering the hidden mapping.
ABSTRACT The high mobility and decentralized nature of Vehicular Ad Hoc Networks (VANETs) present significant security challenges. Specifically, detecting attacks and establishing secure, reliable routing protocols are major critical concerns in the vehicular environment. These attacks can significantly degrade network performance and hinder communication between vehicles. Insider attacks, such as Blackhole attacks, have the potential to severely disrupt VANET systems. This study introduces a novel trust management scheme that incorporates cryptographic techniques to address the important issues of secure routing in VANETs, which also helps in the detection of attacks. In this work, nodes' trust scores are evaluated, and the forwarding node for packet dissemination is chosen based on these scores. Furthermore, an elliptic curve cryptographic (ECC) signcryption technique is added for providing security to the network by authenticating the nodes, which mitigates the misbehaving nodes from the network. The simulation and comparative analysis show the efficacy of the proposed scheme. The proposed approach attained a packet delivery ratio (PDR) of 92.8%, indicating high reliability in data dissemination. Furthermore, the achieved results of throughput and End‐to‐End (E2E) delay are 232.32 KBps and 0.02 s, respectively. The obtained outcomes show enhancements of 94.182%, 49.67%, and 6% in PDR, throughput, and E2E delay, respectively, with respect to the existing techniques.
This paper developed an adaptive federated learning framework for predicting the adaptability levels of students in an online education environment in nonindependent and identically distributed (non-IID) conditions. In practical situations, student data is usually distributed in different multiple institutions, and because of privacy concerns, direct data sharing is often restricted. To address these challenges, the proposed method in this paper introduced an adaptive aggregation mechanism, noise-based privacy enhancement and Top-k model compression techniques. The adaptive aggregation strategy allocates dynamic weights to different clients based on the data size and distribution similarity to reduce the negative impact of heterogeneous data. At the same time, noise-based privacy enhancement ensures data security, while Top-k compression significantly reduces the communication cost during the model transmission process. Experimental results show that compared with centralized learning and FedAvg, the method proposed in this paper is competitive in terms of performance and reduces the communication cost by nearly 60%. Further analysis through confusion matrices and t-SNE visualization shows that the method proposed in this paper improves feature separability and generates more balanced prediction results in different categories. These results suggest that this framework can effectively handle nonindependent and nonidentically distributed data and improve the generalization ability of the model. This study provides a practical and effective solution for collaborative learning in online education to protect privacy. It has potential application value in student performance prediction and personalized learning systems.
Big Data has emerged as a demanding and fast-growing technology in various fields, such as healthcare, finance, banking, education, and the Internet of Things (IoT). Nevertheless, such exponential growth also implies serious challenges and new threats associated with information security, real-time data processing, and system flexibility. These are compounded by the fact that heterogeneous (non-independent and Identically Distributed [IID]) data exist both in IoT and edge clients, and that there are stringent latency and energy constraints on resource-limited devices. Despite various suggested security models, there are still too few to implement full security protection in Big Data settings. This study aims to address these drawbacks by introducing a new Adaptive Hybrid Security Framework that implements federated learning, near-lossless AI-based data compression, reinforcement learning (RL), and a lightweight blockchain with Edge AI for anomaly detection in a single architecture. In contrast to current methods, the framework proposed works by categorizing data dynamically (in accordance with its sensitivity) and dynamically encrypting data according to system load and other security needs, and provides decentralized key management to avoid the presence of single points of failure, which reduces computational requirements. The use of the CICIDS2017 dataset with experimental evaluation showed 94.49% federated classification accuracy, a compression ratio of 3.5-1, a reconstruction loss of 0.004, and an F1-score of 93% for anomaly detection. The system had a latency of 180 ms during encryption and 15 ms during anomaly detection. Overall, the proposed framework provides scalable, efficient, and adaptive security for Big Data, and its lightweight architecture is compatible with real-time usage in IoT and edge setups, which are resource-constrained.
The water and wastewater sector faces growing cyber threats due to rapid digitalization and the use of IoT-based control systems. Many utilities manage essential services that affect public health and the environment but do not have enough cybersecurity staff and cannot afford regular security tests. Traditional penetration testing is often too expensive, complex, and requires expert knowledge. This work presents Red-Pi, a portable and low-cost platform for the cybersecurity assessment of water systems. Red-Pi uses a single-board computer (SBC) and automates main security tasks: reconnaissance, vulnerability detection, exploitation, and report generation. It supports both IT and OT protocols, and it includes encrypted storage, modular connectivity, and secure remote control. Following a six-stage workflow based on MITRE ATT&CK, Red-Pi allows realistic tests in labs and real utility environments. An evaluation in 30 deployments showed a 62.9% reduction in assessment time (Cohen's ), 91.9% cost savings, and 91.4% F1-score for vulnerability detection. With a Composite Performance Index of 84.5/100, Red-Pi helps small and medium utilities perform continuous automated cybersecurity audits, thereby increasing assessment frequency by over 9000% and improving digital resilience according to the NIS2 Directive requirements. This approach reduces barriers and enhances security in critical water infrastructure.
ABSTRACT Elliptic curve cryptography ( ECC ) underpins the security of most blockchain systems, yet its practical implementations face numerous vulnerabilities. In this systematic literature review ( SLR ), we catalogue and analyze attacks on ECC in the context of blockchain security, including side‐channel attacks, nonce/ PRNG failures, cryptanalysis, and implementation flaws, and we survey proposed countermeasures. We follow rigorous SLR methodology with defined inclusion/exclusion criteria, search strategies across databases such as IEEE Xplore, ACM , Scopus, Web of Science, and clear data synthesis, ensuring replicability. Emphasizing empirical case studies and real‐world exploits, we discuss instances where ECC weaknesses led to blockchain breaches including biased elliptic curve digital signature algorithm nonces exposing Bitcoin/Ethereum private keys, smartphone power analysis revealing wallet keys, and Trezor hardware‐wallet key extraction via single‐trace side‐channel analysis ( SCA ). We tabulate known attack vectors versus affected systems, and similarly compare countermeasure techniques such as hybrid classical/quantum schemes, threshold signatures, and zero‐knowledge proofs, along with implementation trade‐offs. We evaluate advances such as Curve25519/ EdDSA and ARM SVE2 to mitigate side‐channel leakage. Our findings highlight that practical security of blockchain cryptosystems depends on correct ECC implementation and emerging cryptographic upgrades, not merely on the mathematical hardness of the elliptic curve discrete logarithm problem.
Cross-Site Scripting (XSS) continues to pose serious risks to modern web applications as attackers increasingly employ sophisticated obfuscation and adversarial manipulations. Traditional rule-based and machine learning detectors often fail to model multi-level semantic patterns, making them vulnerable to well-crafted evasive payloads. To address this challenge, this paper proposes an adversarially robust CNN-BiLSTM-Multi-Head Attention model that enhances contextual feature extraction and strengthens resistance to gradient-driven perturbations. Convolutional layers capture local token interactions, BiLSTM modules learn long-range dependencies, and the attention mechanism provides token-level interpretability and prioritized feature weighting. FGSM-based adversarial training further exposes the model to worst-case perturbation directions, improving its generalization to evolving XSS variants. Experiments conducted on a composite dataset of 29 000 real-world, benchmark, and adversarial samples demonstrate the effectiveness of the proposed framework. The system achieves 97.8% accuracy, 96.5% precision, 97.1% recall, and a 98.3% AUC score, surpassing existing detectors such as FusionXSS and XSShield. Robustness evaluations show gains of up to 17% under FGSM attacks. With low inference latency and moderate memory usage, the model offers a practical and scalable solution for real-time XSS defense in modern web environments.
Intrusion detection systems (IDS) are critical to the security of the dynamic internet of things (IoT) environment. The integration of Artificial Intelligence (AI) into IDS has substantially improved network security. Particularly, deep learning techniques have shown strong potential in addressing IoT security challenges. However, the effectiveness and generalizability of these methods are highly dependent on the quality, diversity, and representativeness of the input data. Inadequate data can significantly impact model training, resulting in biased predictions, reduced accuracy, and vulnerabilities in real-world deployments. This research aims to extract meaningful, informative, and discriminative features from the data to enhance the quality of input for deep learning models, thereby improving their learning efficiency, accuracy, and overall performance. A novel genetic algorithm-based artificial neural network (GA-ANN) is introduced with an advanced feature extraction method that optimizes feature vectors for various configurations of the ANN. The study thoroughly evaluates feature selection techniques for their impact on IDS performance. Selecting only relevant features, rather than including all of them, significantly reduces the size of the feature set. This optimization enhances both efficiency and detection capability by lowering computational overhead while improving accuracy. On the CIC2023 dataset containing 34 attack classes, the GA-ANN model achieved 98.46% accuracy, surpassing conventional techniques that were limited to smaller datasets with up to 11 classes. The study contributes by tackling shortcomings in data selection quality and strengthening IDS effectiveness in dynamic and rapidly evolving IoT environments.
In the highly interconnected industrial internet, security threats often exhibit multi-stage and cross-domain cascading characteristics. Traditional defense strategies that depend on static configurations are insufficient to support real-time and anticipatory protection. To address this limitation, this paper proposes an industrial situation-prediction-driven role-switching stochastic Stackelberg game method. First, we construct an industrial network situation prediction model constrained by attack graphs. By incorporating a reachability-mask matrix and vulnerability exposure metrics, the proposed model enables interpretable and forward-looking situation prediction while strictly preventing illegal state transitions. Then, we propose a role-switching stochastic Stackelberg game framework that integrates leader-follower role transitions induced by real-time and predicted situations into a unified decision-making model. Finally, we introduce a situation-prediction-driven multi-agent reinforcement learning algorithm to approximate game equilibria, enabling efficient computation of optimal pre-deployment defense strategies. Experimental results on a simulation testbed demonstrate that the proposed method improves prediction accuracy by approximately 11% over baseline approaches and completely eliminates unreachable state transitions. The prediction-enhanced adaptive defense strategy reduces attack success rates by 57%, decreases critical asset loss by about 48%, and significantly extends the attack chain length, yielding a defense strategy that is both cost-efficient and loss-minimizing.
At present, traditional homomorphic encryption (HE) algorithms face the problems of high computational consumption and weak antiattack ability in complex environments. In order to handle these shortcomings and improve the HE algorithm, the ring learning with errors (RLWE) is proposed. The results show that the proposed algorithm can effectively reduce the computational overhead by combining local differential privacy mechanisms with Nth-degree truncated polynomial ring units (NTRU) lattice-based lightweight encryption schemes. Additionally, data security can be made better. On the Espressif Systems Protocol 32-bit (ESP32) platform, encryption latency is reduced by 42.1% compared to the Cheon-Kim-Kim-Song (CKKS) algorithm. With the incorporation of an adaptive mechanism, the attack rate of breaking through the defense line is reduced to below 12%. Meanwhile, this algorithm can also guarantee the quantum-resistant security strength of 148 bits in the configuration of 2048-bit security parameters; it comprehensively surpasses CKKS, RLWE-CKKS, and many improved baseline algorithms. Under the harsh conditions of the Internet of Things (IoT) terminal, this study simultaneously tackles computational constraints, privacy leakage risks, and postquantum threats; thus, it can offer a theoretical reference and engineering support for building a trustworthy IoT system.
Healthcare supply chains and IoMT-based clinical monitoring depend on continuous data sharing but remain at risk of attacks, counterfeit drugs, and privacy breaches. Although federated learning helps protect data, it still faces challenges such as malicious clients, spoofing, and inconsistent participation. To overcome these issues, this paper introduces an Adaptive Trust-Driven Blockchain Federated Learning (ATB-FL) framework. The design combines behavior-based trust assessment, blockchain-enabled authentication, and incentive-penalty strategies within a scalable security model. This approach provides tamper-proof traceability, real-time participant validation, and compliance with regulatory standards while keeping sensitive data local. Using the CIC-IoMT 2024 dataset, ATB-FL achieved 95.1% diagnostic accuracy, reduced misclassification to under 5%, and increased blockchain throughput by over 30% compared with existing methods. The effectiveness of the framework is further illustrated through case studies on vaccine cold-chain monitoring and remote diagnostics. Overall, ATB-FL offers a reliable and practical foundation for building secure and privacy-preserving healthcare systems of the future.
With the digital transformation of industrial enterprises, Network Security Situation (NSS) assessment and prediction have become key tasks to ensure enterprise information security. To evaluate and predict the NSS more accurately, this study proposes an SAE-DBSCAN algorithm and an improved Bi-LSTM. Research showed that the ACA of the SAE-DBSCAN on different data was 91.3% and 90.7%, which was numerically better than the comparative models. The precision and recall rates after improved Bi-LSTM convergence were 0.977 and 0.938. When this model was applied to different NSSs, its predicted value was closest to the real value. Especially in the DoS attack type, the difference between the predicted and real values was only 0.008. The outcomes demonstrate that SAE-DBSCAN can effectively assess the NSS, the improved Bi-LSTM's performance is better, and the prediction ability is better. This study can help enterprises more accurately identify and respond to potential network security threats, thereby optimizing resource allocation and reducing security risks.
The rising rate of cyberattacks and the dearth of skilled cybersecurity professionals require innovative solutions beyond traditional security measures. Large Language Models (LLMs), known for their natural language processing abilities, present a viable way to improve cybersecurity defenses. However, despite LLMs' exploration in various security applications, systematic research aligning LLMs' contributions with established cybersecurity frameworks is lacking. To bridge this gap, this paper presents a systematic literature review aligned with the NIST Cybersecurity Framework (CSF 2.0) to gain a clear understanding of LLMs' multifaceted contributions and uncover their overlooked potential, particularly in areas such as the Awareness and Training category of the NIST CSF 2.0 Protect function. Leveraging the accessibility and privacy benefits of open-source LLMs (OSLLMs), we developed a benchmarking methodology using Multiple-Choice Question Answering (MCQA) to evaluate 21 OSLLMs recognized for their state-of-the-art performance (e.g., Llama-3-70B and Qwen2-72B, over 80% in MMLU) across two publicly available datasets. Our findings reveal that while larger LLMs generally outperformed smaller ones, medium-sized LLMs (7B-27B) achieved accuracy within 5-9 percentage points of 70B+ models, contesting the claim that model size alone dictates efficacy. These insights motivated a further investigation into optimized prompt engineering workflows using the DSPy framework. Our results reveal that advanced prompting techniques significantly boost the performance of smaller models, narrowing the performance gap with larger ones and broadening deployment possibilities. Furthermore, we introduced structural modifications and novel exit options to address position and forced-choice biases in standard MCQA datasets, further enhancing the reliability of OSLLM evaluations. Overall, the results of this study not only strengthen our understanding of OSLLMs' aptitude in cybersecurity but also emphasize the impact of advanced prompt engineering and unbiased datasets in enhancing the capabilities of OSLLMs in cybersecurity.
Existing IoT frameworks that rely on traditional blockchain require high computational power and cannot scale effectively in an IoT environment. We introduce Tangle, a DAG-based blockchain framework that uses MAM to implement role-based data access control for cognitive health monitoring. In our proposed approach, we send IoT data to a blockchain network. We added two new nodes in the network, Upload nodes and Processing nodes. Upload nodes encrypt the data and generate the MAM root, which then gets published to the Tangle network. Processing nodes receive this data and decrypt it using their assigned access keys before applying preprocessing steps and running Machine Learning algorithms. To handle both static and dynamic IoT data features effectively, we designed a transformer-based incremental ML model for training. In our experimental results, our model attains a 98% accuracy compared to several existing state-of-the-art machine learning techniques.
Leading enterprises typically choose for a centralized, large-scale database storage system. Data security is crucial for communication and database preservation. The principal issue is governing unauthorized access to the database. Role-based access control (RBAC) is a recognized approach for maintaining confidentiality in database systems. Nevertheless, System Administrators (SA) and Database Administrators (DBA) can obtain critical information even within the RBAC architecture. An approach based on encryption has been developed to address this issue. The primary concern is the secure distribution of keys among individuals with shared data access and a group of individuals needing access to shared data, viewed as having a common objective. We have instituted row-level security (RLS) in database design to establish safe access control, called Encrypted Role-Based Access Control (ERBAC). Additionally, we deployed Enterprise Resource Planning (ERP) systems, which are essential applications where security concerns are critical. ERBAC integrates effortlessly with an operational ERP system. We offer security, performance analysis, and encrypted query processing for role-based encrypted data storage when the database contains identical data in many forms.
To address the challenges of real-time control over the dynamic evolution of malware behavior and abnormal traffic in smart grids, this study proposed a dynamic defense model integrating graph attention network (GAT), long short-term memory (LSTM), and adversarial reinforcement learning (ARL). By incorporating the ARL mechanism to optimize the decision-making process of GAT-LSTM, the model extracted more robust spatio-temporal feature representations in adversarial environments, enabling precise prediction of malicious activities. Furthermore, a constraint-optimized active traffic shaping strategy generator was designed to map predicted probabilities in real time to optimal control commands. The study first analyzed the performance of the combination algorithm. The results showed that the algorithm's feature recognition accuracy could reach 96.3%, demonstrating excellent performance. The effectiveness of the malware behavior prediction model was then analyzed. The results showed that the model's prediction accuracy could reach 96.7%. The active traffic shaping model could defend against more than 94% of network attacks. In summary, the dynamic defense method proposed in this study can accurately predict malicious software behavior and defend against most network attacks, thereby ensuring the security of smart grids.
To resolve the conflict between privacy protection and compliance verification of commercial contracts in the digital economy, this study aims to propose the application of a commercial contract generation technology based on verifiable encryption signature algorithms in civil and commercial law. This technology innovatively combines Attribute-Based Encryption (ABE) with Zero-Knowledge Succinct Non-Interactive Argument of Knowledge (zk-SNARKs) to construct a new electronic contract framework. Methodologically, ABE is used to encrypt contract clauses, enabling refined "on-demand decryption" based on the roles and permissions of participants. Meanwhile, zk-SNARKs are utilized to generate proofs for clauses requiring verification (e.g., transaction amount), allowing third parties (such as regulatory authorities) to complete "privacy-preserving verification" without accessing the original contract text. Experimental results show that when the model processes a contract containing 500 clauses, the size of the generated data package is only 64.7 KB, which is approximately 3.8 times smaller than that of the pure ABE encryption model (247.8 KB) and 7.4 times smaller than that of the traditional digital signature model (480.1 KB). At the same time, the key size for users to access 500 clauses is only 5.2 KB, which is much smaller than that of the baseline models. Therefore, this technical solution effectively resolves the long-standing "privacy-verification" dilemma in commercial activities. This also proves its feasibility and economy in practical applications with the advantages of low storage and transmission costs, providing secure and reliable technical support for civil and commercial legal practices in the digital era.
In medical domains, with the rapid development of internet activities, medical professionals have widely used medical information. Secure management and transmission of medical images, which is crucial to facilitate association, while protecting sensitive patient details. Nowadays, the growth of digitalization is providing convenience and competence that have expressively sensitive risks related to cybersecurity, which makes the security of medical information more significant than ever. However, healthcare application content can be easily stored, tampered with, and shared by unauthorized users. Consider these limitations; the concept of watermarking is to conceal secret systems while maintaining significant features like visual quality and robustness. But, the traditional watermarking process has several limitations, such as computational cost, security concerns, and low embedding ability. Here, this research framework introduces a deep learning (DL) based secure and robust image watermarking approach that is combined with chaotic encryption methods. Initially, this model utilizes the adaptive osprey optimization algorithm (AOOA) for selecting optimal locations in cover images. Following this, a chaotic-based encryption process is used to encrypt secret images, which increases security performance. Proposed model presents depthwise separable convolutional assisted generative adversarial network used for embedding secret images into cover images. This model includes a depthwise separable convolution layer to decrease computational complexity and a GAN model for increasing the ability of the watermarking process that enhances visual quality. Experimental analysis, proposed model achieves a PSNR value of 68.9754 dB, a SSIM value of 0.9925, and an accuracy value of 99.45%, which are compared with existing models and state-of-the-art models to demonstrate performance evaluation and robustness, security as well as invisibility performances in the medical image watermarking process.
Intrusion detection systems are vital cybersecurity systems in the modern world. IoT devices have taken over systems around us. Due to the limited computational capability coupled with the open endpoint nature of IoT devices, they are susceptible to attacks. Such attacks jeopardize the security of every device on the network, including vital servers and databases. In this work, we survey and experiment with some famous and widely used machine learning algorithms to detect and mitigate these attacks. We employ a divide and conquer approach to classify the incoming traffic into different types of attacks to counter the skewness in the dataset used. We use the results to evaluate the models on markers like precision, recall, accuracy, and balanced accuracy. The observations are tabulated and concluded based on computational costs and their efficiency. We study the effectiveness of modern machine learning algorithms like decision trees, rotation trees, K-means, ridge, hidden Markov models, gradient boosting, and multiple bagging techniques in differentiating among incoming network traffic.
With the rapid expansion of smart city infrastructure, its highly interconnected IoT devices, sensor networks and key information systems constitute a huge and complex digital ecosystem that is extremely vulnerable to various cyber threats. The forms of network attacks are becoming increasingly intelligent and concealed. It is challenging to address unknown threats effectively and rapidly evolving attack patterns within massive, multi-source, heterogeneous data streams. Traditional methods are difficult to deal with threats, which in turn pose a threat to the stability of urban operations and other aspects. This study proposes an innovative predictive deep learning framework, which aims to achieve early detection and accurate prediction of potential threats in smart city environments. The model integrates a spatio-temporal graph neural network and a self-attention mechanism. This combination deeply mines the complex correlation patterns hidden among multidimensional data, such as massive network equipment logs, traffic control system operating status, and public Wi-Fi access traffic, and infers possible future attack trends in real time. Rigorously tested in a high-fidelity simulation environment containing 1200,000 real smart city network equipment logs and 20 typical threat activities, the model demonstrated excellent performance. It achieved a threat detection accuracy rate of 96.7% and an F1 score of 97.2%, which is significantly better than the current mainstream detection methods. What is particularly outstanding is its prediction ability. The model's prediction accuracy rate for new attack samples that are unknown reaches 91.5%. In terms of real-time response, the average detection delay is controlled within 0.12 s, verifying the feasibility and efficiency of its deployment in large-scale, dynamic, and complex network environments. It provides core support technology for active security protection in smart cities.
