Security Vulnerabilities Research Articles

Secure and Lightweight Firmware Over-the-Air Update Mechanism for Internet of Things

The Internet of Things (IoT) necessitates secure and lightweight firmware over-the-air (FOTA) update mechanisms for remote device management and timely mitigation of security vulnerabilities. This study introduces an FOTA update method to mitigate man-in-the-middle attacks in resource-constrained environments. The proposed method minimizes firmware file size and encryption overhead through a dual-XOR operation and DEFLATE compression, while enhancing security via multiple transmission channels. It improves performance in terms of latency, memory usage, and power consumption but also maintains security against brute-force attacks during MITM attacks.

APIRL: Deep Reinforcement Learning for REST API Fuzzing

REST APIs have become key components of web services. However, they often contain logic flaws resulting in server side errors or security vulnerabilities. HTTP requests are used as test cases to find and mitigate such issues. Existing methods to modify requests, including those using deep learning, suffer from limited performance and precision, relying on undirected search or making limited usage of the contextual information. In this paper we propose APIRL, a fully automated deep reinforcement learning tool for testing REST APIs. A key novelty of our approach is the use of feedback from a transformer module pre-trained on JSON-structured data, akin to that used in API responses. This allows APIRL to learn the subtleties relating to test outcomes, and generalise to unseen API endpoints. We show APIRL can find significantly more bugs than the state-of-the-art in real world REST APIs while minimising the number of required test cases. We also study how reward functions, and other key design choices, affect learnt policies with a thorough ablation study.

Medical MLLM Is Vulnerable: Cross-Modality Jailbreak and Mismatched Attacks on Medical Multimodal Large Language Models

Security concerns related to Large Language Models (LLMs) have been extensively explored; however, the safety implications for Multimodal Large Language Models (MLLMs), particularly in medical contexts (MedMLLMs), remain inadequately addressed. This paper investigates the security vulnerabilities of MedMLLMs, focusing on their deployment in clinical environments where the accuracy and relevance of question-and-answer interactions are crucial for addressing complex medical challenges. We introduce and redefine two attack types: mismatched malicious attack (2M-attack) and optimized mismatched malicious attack (O2M-attack), by integrating existing clinical data with atypical natural phenomena. Using the comprehensive 3MAD dataset that we developed, which spans a diverse range of medical imaging modalities and adverse medical scenarios, we performed an in-depth analysis and proposed the MCM optimization method. This approach significantly improves the attack success rate against MedMLLMs. Our evaluations, which include white-box attacks on LLaVA-Med and transfer (black-box) attacks on four other SOTA models, reveal that even MedMLLMs designed with advanced security mechanisms remain vulnerable to breaches. This study highlights the critical need for robust security measures to enhance the safety and reliability of open-source MedMLLMs, especially in light of the potential impact of jailbreak attacks and other malicious exploits in clinical applications. Warning: Medical jailbreaking may generate content that includes unverified diagnoses and treatment recommendations. Always consult professional medical advice.

Coding Bugs Leading to Security Vulnerabilities in Windows Drivers

Coding Bugs Leading to Security Vulnerabilities in Windows Drivers

Against All Odds: Overcoming Typology, Script, and Language Confusion in Multilingual Embedding Inversion Attacks

Large Language Models (LLMs) are susceptible to malicious influence by cyber attackers through intrusions such as adversarial, backdoor, and embedding inversion attacks. In response, the burgeoning field of LLM Security aims to study and defend against such threats. Thus far, the majority of works in this area have focused on monolingual English models; however, emerging research suggests that multilingual LLMs may be more vulnerable to various attacks than their monolingual counterparts. While previous work has investigated embedding inversion over a small subset of European languages, it is challenging to extrapolate these findings to languages from different linguistic families and with differing scripts. To this end, we explore the security of multilingual LLMs in the context of embedding inversion attacks and investigate cross-lingual and cross-script inversion across 20 languages, spanning over 8 language families and 12 scripts. Our findings indicate that languages written in Arabic and Cyrillic scripts are particularly vulnerable to embedding inversion, as are languages within the Indo-Aryan language family. We further observe that inversion models tend to suffer from language confusion, sometimes significantly reducing the efficacy of an attack. Accordingly, we systematically explore this bottleneck for inversion models, uncovering predictable patterns attackers could leverage. Ultimately, this study aims to further the field's understanding of the outstanding security vulnerabilities facing multilingual LLMs and raise awareness for the languages most at risk of negative impact from these attacks.

Enhancing Blood Supply Chain Management with Blockchain Technology to Improve Diagnostic Precision and Strengthen Health Information Security

The efficient management of blood supply chains is critical to ensuring timely availability, diagnostic precision, and overall healthcare reliability. However, traditional blood supply chain systems face challenges such as data inaccuracies, lack of transparency, and vulnerabilities in health information security. This review explores the integration of blockchain technology to enhance the efficiency, traceability, and security of blood supply chain management. Blockchain’s decentralized and immutable ledger enables real-time tracking of blood donations, storage conditions, and transfusion records, reducing risks of contamination, fraud, and mismanagement. By leveraging smart contracts and cryptographic security, blockchain enhances interoperability between healthcare providers, regulatory bodies, and blood banks while ensuring compliance with data privacy regulations. Additionally, blockchain-powered solutions improve diagnostic precision by integrating secure, tamper-proof patient records, facilitating seamless data exchange for transfusion compatibility and disease screening. This paper examines case studies, emerging blockchain frameworks, and key adoption challenges to highlight the transformative impact of blockchain in optimizing blood supply chains. The findings suggest that blockchain adoption can significantly enhance supply chain efficiency, improve patient outcomes, and strengthen health information security, paving the way for a more transparent and resilient healthcare ecosystem.

A systematic review on smart contracts security design patterns

Smart contracts have accelerated the adoption of blockchain technology across various domains by enabling coded agreements between transaction participants. However, increased software defects and vulnerabilities in smart contracts, driven by developer inexperience with languages like Solidity and a lack of effective detection tools, pose significant risks. Given the high value of assets managed on blockchain (e.g., cryptocurrencies), these vulnerabilities can lead to severe consequences. Researchers and practitioners have proposed numerous smart contract design patterns to mitigate certain faults or vulnerabilities. Despite these efforts, it remains unclear which types of defects these patterns target and how effectively they address the wide range of existing smart contract security vulnerabilities. In this paper, we review the state of the art in smart contract design patterns, categorizing them and analyzing their effectiveness in mitigating known security vulnerabilities. Our findings reveal that only five patterns directly aim to prevent security vulnerabilities, collectively addressing just 6 out of 94 security issues identified by OpenSCV (a state-of-the-art vulnerability taxonomy), highlighting the need for further research on smart contract security design patterns.

FRACE: Front-Running Attack Classifcation on Ethereum using Ensemble Learning

Abstract With the rapid evolution of blockchain technologies, Ethereum has emerged as a central platform for advanced financial applications but has concurrently experienced a rise in security vulnerabilities, particularly from front-running attacks. These attacks exploit transaction sequencing for illegal gains. To combat this, we introduce FRACE (Front-Running Attack Classification using Ensemble Learning), a novel methodology that classifies front-running attacks into displacement, insertion, and suppression using an ensemble learning model. This precise classification facilitates tailored defensive strategies, enhancing the robustness and accuracy of attack detection. Our approach achieves an accuracy of 95.36% and an F1-score of 95.30%, significantly improving the security of decentralized applications. Extensive analysis and validation on Ethereum confirm these results. Future efforts will refine these models and extend their application to other blockchain platforms, striving for a universally secure, transparent, and reliable digital transaction ecosystem.

A deep Reinforcement learning-based robust Intrusion Detection System for securing IoMT Healthcare Networks

The Internet of Medical Things (IoMT) is transforming healthcare by enabling continuous remote patient monitoring, diagnostics, and personalized therapies. However, the widespread deployment of these devices introduces significant security vulnerabilities due to limited resources and inadequate network protocols. Intrusions within IoMT networks can compromise patient privacy, disrupt critical medical services, and jeopardize patient safety. To address these challenges, we propose HCLR-IDS, an advanced Intrusion Detection System (IDS) specifically designed for IoMT networks. The system integrates Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and Reinforcement Learning (RL) techniques, namely Deep Q-Network (DQN) and Proximal Policy Optimization (PPO), to enhance the detection of evolving threats. The methodology begins with Enhanced Mutual Information Feature Selection (MIFS) to preprocess the CICIoMT2024 dataset, selecting the most relevant features while reducing noise and computational complexity. These selected features are then passed through a hybrid CNN-LSTM architecture. The CNN captures spatial patterns in network traffic, while the LSTM identifies temporal patterns. This dual feature extraction approach enables the system to effectively detect both static and dynamic characteristics of IoMT data. After feature extraction, the model incorporates DQN and PPO for decision-making. DQN optimizes actions based on Q-values, enhancing detection rewards, while PPO ensures stability in dynamic environments through a clipping mechanism. This combination of adaptive Q-learning and stable policy optimization significantly improves system robustness, ensuring effective real-time intrusion detection. The model demonstrates exceptional performance with binary classification accuracy of 0.9958, outperforming traditional IDS models. Additionally, it performs effectively in multi-class classification across 18 classes, achieving an accuracy of 0.7773. These results highlight that HCLR-IDS offers a reliable and efficient solution for securing IoMT healthcare systems.

Advancing clinical biochemistry: addressing gaps and driving future innovations

Modern healthcare depends fundamentally on clinical biochemistry for disease diagnosis and therapeutic guidance. The discipline encounters operational constraints, including sampling inefficiencies, precision limitations, and expansion difficulties. Recent advancements in established technologies, such as mass spectrometry and the development of high-throughput screening and point-of-care technologies, are revolutionizing the industry. Modern biosensor technology and wearable monitors facilitate continuous health tracking, Artificial Intelligence (AI)/machine learning (ML) applications enhance analytical capabilities, generating predictive insights for individualized treatment protocols. However, concerns regarding algorithmic bias, data privacy, lack of transparency in decision-making (“black box” models), and over-reliance on automated systems pose significant challenges that must be addressed for responsible AI integration. However, significant limitations remain—substantial implementation expenses, system incompatibility issues, and information security vulnerabilities intersect with ethical considerations regarding algorithmic fairness and protected health information. Addressing these challenges demands coordinated efforts between clinicians, scientists, and technical specialists. This review discusses current challenges in clinical biochemistry, explicitly addressing the limitations of reference intervals and barriers to implementing innovative biomarkers in medical settings. The discussion evaluates how advanced technologies and multidisciplinary collaboration can overcome these constraints while identifying research priorities to enhance diagnostic precision and accessibility for better healthcare delivery.

Integration of Quantum Key Distribution in 6G Passive WDM Optical Fronthaul Architecture

The advancement of 6G networks demands high-capacity, ultra-low latency, and secure communication infrastructure. Traditional fronthaul solutions face significant challenges in meeting these demands due to increasing data traffic, network densification, and security vulnerabilities. To address these issues, this paper proposes an innovative 6G passive Wavelength Division Multiplexing (WDM) fronthaul architecture integrated with Quantum Key Distribution (QKD). By leveraging quantum cryptographic principles, our approach ensures unbreakable security and protects network transmissions from potential cyber threats and eavesdropping attacks. The proposed architecture effectively combines high-speed optical transmission with quantum-secured key exchange, optimizing network reliability and reducing security risks. Through comprehensive simulations, we evaluate key performance factors, including secure key rate (SKR), quantum bit error rate (QBER), detector efficiency, and quantum transmission loss. Results indicate that our architecture maintains high SKR with minimal QBER, even under challenging transmission conditions, ensuring robust quantum-secured communication. Additionally, the passive nature of WDM significantly reduces power consumption and maintenance costs, improving network sustainability. This research establishes a practical pathway for integrating quantum security into next-generation optical fronthaul networks, enabling highly secure, scalable, and efficient data transmission for future 6G applications, including autonomous vehicles, smart cities, and industrial IoT.

Cloud at the Crossroads: How Secure Integrations Shape Education, Law Enforcement, and Art

Cloud at the Crossroads: How Secure Integrations Shape Education, Law Enforcement, and Art examinesthe transformative impact of Enterprise Cloud Architecture across divergent domains. This article revealshow cloud computing technologies fundamentally alter established paradigms through enhanced cross-platform integration capabilities and robust security frameworks. Integrating cloud-based systems ineducational environments has expanded access to personalized learning experiences while presentingsignificant data privacy challenges. In law enforcement, cloud platforms have revolutionized evidencemanagement and introduced predictive policing capabilities that enhance public safety but raise complexethical questions. The artistic domain has witnessed unprecedented democratization through cloud-basedcreative tools and blockchain-integrated distribution platforms that redefine traditional concepts ofcreativity and ownership. These cross-domain implementations share common ethical challenges, includingdata privacy concerns, security vulnerabilities, algorithmic bias, and digital divide considerations.Addressing these challenges requires balanced approaches incorporating comprehensive securityframeworks, transparent governance structures, and equitable implementation strategies. As cloudtechnologies evolve, understanding these transformations becomes increasingly vital for stakeholdersseeking to harness their potential while mitigating associated risks.

Generalized Singular Value Decomposition-Based Secure Beam Hybrid Precoding for Millimeter Wave Massive Multiple-Input Multiple-Output Systems

The precoder obtained using the traditional singular value decomposition (SVD) method for legitimate user’s channel, while achieving the highest spectral efficiency for the legitimate user, cannot defend against eavesdropping attacks, thus posing a security vulnerability. This paper investigates the millimeter wave (mmWave) secure beam hybrid precoding technology and proposes a generalized singular value decomposition (GSVD)-based secure beam hybrid precoding algorithm, termed GSVD-Sparsity, leveraging the sparsity of the mmWave beamspace channel. The algorithm selects the most powerful paths from the legitimate user’s beamspace channel representation and utilizes their corresponding angle information to construct a radio frequency (RF) precoder. It then constructs a hybrid precoder that closely approximates the optimal digital precoder derived from the GSVD-based scheme in a fully digital system. The simulation results indicate that, compared to the SVD-based scheme that focuses on spectral efficiency, the GSVD-based precoding scheme can form secure beams in a fully digital system. Under the condition that the legitimate user experiences a certain loss in the received signal-to-noise ratio (SNR), the eavesdropper is unable to correctly reconstruct the original constellation diagram, ensuring the scheme has strong anti-eavesdropping capabilities. In a hybrid precoding system, the low-complexity GSVD-Sparsity algorithm can achieve a spectral efficiency close to that of the GSVD-based scheme in a fully digital system while maintaining anti-eavesdropping capabilities.

On the effects of program slicing for vulnerability detection during code inspection

Slicing is a fault localization technique that has been proposed to support debugging and program comprehension. Yet, its empirical effectiveness during code inspection by humans has received limited attention. The goal of our study is two-fold. First, we aim to define what it means for a code reviewer to identify the vulnerable lines correctly. Second, we investigate whether reducing the number of to-be-inspected lines by method-level slicing supports code reviewers in detecting security vulnerabilities. We propose a novel approach based on the notion of a δ-neighborhood (intuitively based on the idea of the context size of the command git diff) to define correctly identified lines. Then, we conducted a multi-year controlled experiment (2017-2023) in which MSc students attending security courses (n=236) were tasked with identifying vulnerable lines in original or sliced Java files from Apache Tomcat. We provide perfect seed lines for a slicing algorithm to control for confounding factors. Each treatment differs in the pair (Vulnerability, Original/Sliced) with a balanced design with vulnerabilities from the OWASP Top 10 2017: A1 (Injection), A5 (Broken Access Control), A6 (Security Misconfiguration), and A7 (Cross-Site Scripting). To generate smaller slices for human consumption, we used a variant of intra-procedural thin slicing. We report the results for δ=0 which corresponds to exactly matching the vulnerable ground truth lines, and δ=3 which represents the scenario of identifying the vulnerable area. For both cases, we found that slicing helps in ‘finding something’ (the participant has found at least some vulnerable lines) as opposed to ‘finding nothing’. For the case of δ=0 analyzing a slice and analyzing the original file are statistically equivalent from the perspective of lines found by those who found something. With δ=3 slicing helps to find more vulnerabilities compared to analyzing an original file, as we would normally expect. Given the type of population, additional experiments are necessary to be generalized to experienced developers.

Neuro-Symbolic AI for IoT-Driven Smart Cities: A Next-Generation Framework for Urban Intelligence

The integration of the Internet of Things (IoT) and Artificial Intelligence (AI) is revolutionizing urban landscapes by enhancing operational efficiency, resource management, and sustainability in smart cities. IoT enables real-time data acquisition through distributed sensor networks, while AI processes this data to facilitate intelligent decision-making across critical urban domains, including transportation, energy management, environmental monitoring, public safety, and healthcare. Despite its potential, this convergence presents critical challenges such as data heterogeneity, security vulnerabilities, computational constraints, and regulatory compliance. This paper provides a comprehensive review of the opportunities presented by IoT-AI integration, analyzing key enabling technologies such as edge computing, federated learning, and privacy-preserving AI models. The study further examines major challenges, including interoperability constraints, security risks, and ethical considerations, while exploring advanced mitigation strategies such as blockchain-enhanced security, decentralized intelligence, and adaptive AI-driven urban systems. Additionally, this paper outlines future prospects, focusing on the transformative role of 5G, digital twins, and quantum computing in next-generation smart cities. By synthesizing recent advancements and addressing critical research gaps, this study offers valuable insights for researchers, policymakers, and urban planners striving to build resilient, scalable, and sustainable smart city ecosystems.

A hybrid blockchain-enabled payment system for efficient electronic vehicle charging payments on electrified roads

A hybrid blockchain-enabled payment system for efficient electronic vehicle charging payments on electrified roads

An IPv6 address fast scanning method based on local domain name association

With the increase of security issues in IPv6 networks, conducting address scanning in IPv6 networks proves beneficial for identifying potential security risks and vulnerabilities. To enhance the privacy of users’ IPv6 addresses, mainstream OS (Operating System) nodes currently employ randomized interface identifiers and temporary IPv6 addresses. Additionally, since most existing IPv6 address scanning methods rely on active scanning, which makes current on-link IPv6 address scanning methods face the challenges of incomplete scan results, poor coverage across different OSs, significant impact on network performance, and the inability to promptly detect subsequently joined hosts. To this end, An IPv6 address fast scanning method based on local domain name association (FScan6), which combines active scanning and passive listening, is proposed. The active scanning module targets different OSs using distinct protocols (Browser and DNS-SD) to obtain local domain names of on-link hosts. Meanwhile, the passive listening module monitors traffic to extract local domain names of on-link hosts. Then, it employs mDNS protocol to retrieve IPv6 addresses associated with these local domain names. A typical on-link IPv6 network environment was constructed, comprising 26 versions of Windows, Apple, and Linux OSs, and FScan6 was compared with 9 IPv6 address scanning methods. The experimental results show that FScan6 outperforms existing IPv6 address scanning methods in terms of OS coverage and scanning result completeness. Specifically, regarding OS coverage, FScan6 successfully detected all IPv6 addresses across 26 different OS versions, which outperformed 9 address scanning tools and scripts by a factor of 2.89 times at most. Regarding scanning result completeness, FScan6 identified up to 54 additional IPv6 addresses at most compared to these tools and scripts. Additionally, FScan6 has a minimal impact on network performance, with the packet loss rate induced by the tool consistently remaining at 0%.

Optimization of Serverless Mobile Cloud Applications for Enhanced Security and Resource Efficiency

Serverless mobile cloud applications, while highly scalable and cost-efficient, face critical challenges in terms of security vulnerabilities and inefficient resource management. These challenges are amplified by the dynamic nature of mobile environments and the reliance on third-party infrastructure. This research specifically addresses the problem of securing data transmission, managing authentication and optimizing the allocation of computing resources in serverless mobile cloud architectures. To resolve these issues, the paper proposes a novel, integrated framework that leverages machine learning to predict application resource demands and proactively scale serverless functions. The approach includes the implementation of multi-factor authentication, role-based access control, and encryption protocols to enhance data confidentiality and system integrity. This research demonstrates significant improvements in application performance, reduced infrastructure costs, better cache management, and a notable reduction in latency and miss rates. The proposed model enhances serverless mobile cloud applications' reliability, scalability, and cost-effectiveness by combining adaptive security mechanisms with intelligent resource provisioning

An IoT Featureless Vulnerability Detection and Mitigation Platform

With the increase in ownership of Internet of Things (IoT) devices, there is a bigger demand for stronger implementation of security mechanisms and addressing zero-day vulnerabilities. This work is the first to provide a platform that combines featureless approaches with artificial intelligence (AI) algorithms, which are deep learning and large language models, to uncover IoT security vulnerabilities based on network traffic data directly without manual feature selection. The platform correctly identifies vulnerable and secure IoT devices just by raw network traffic! Experimental results show that the proposed study detects vulnerability with great accuracy by using pre-trained deep learning and LLM models, which facilitates direct extraction of vulnerability features from the dataset and therefore helps speed up the identification process. In addition, the design of the platform ensures that the models are accessible and can be easily applied by users with a user-friendly interface. Furthermore, the models with small sizes, 277.5 MB and 334 MB for the deep learning model and the LLM model, respectively, illustrated the potential use of the detection tool in practical settings. The ability to defend large-scale, diversified IoT ecosystems efficiently and in a scalable way by installing thousands of software manifestations quickly while exposing new applications to growing cyber threats is made possible by this significant advancement in the field of IoT security.

Blockchain-enhanced vehicle data management system with file-based search for crime detection

Effective vehicle data management is crucial for traffic safety and public security, requiring both rapid processing and robust security mechanisms. This study presents a hybrid system that combines a file-based search algorithm with blockchain technology to improve performance and security. The file-based search algorithm facilitates quick data retrieval, while blockchain ensures data integrity and protects against tampering. By integrating these two methods, the system overcomes common bottlenecks and security vulnerabilities found in centralized architectures. Experimental results reveal that the proposed system enhances data processing speed by 84%, considerably increasing data security. These findings highlight its practical applications in smart cities, traffic management, and public safety.