Sedat Akleylek’s research while affiliated with University of Tartu and other places

Uncrewed Aerial Vehicles (UAVs) are frequently utilized in several domains such as transportation, distribution, monitoring, and aviation. A significant security vulnerability is the Global Positioning System (GPS) Spoofing attack, wherein the assailant deceives the GPS receiver by transmitting counterfeit signals, thereby gaining control of the UAV. This can result in the UAV being captured or, in certain instances, destroyed. Numerous strategies have been presented to identify counterfeit GPS signals. Although there have been notable advancements in machine learning (ML) for detecting GPS spoofing attacks, there are still challenges and limitations in the current state-of-the-art research. These include imbalanced datasets, sub-optimal feature selection, and the accuracy of attack detection in resource-constrained environments. The proposed framework investigates the optimal pairing of feature selection (FS) methodologies and deep learning techniques for detecting GPS spoofing attacks on UAVs. The primary objective of this study is to address the challenges associated with detecting GPS spoofing attempts in UAVs. The study focuses on tackling the issue of imbalanced datasets by implementing rigorous oversampling techniques. To do this, a comprehensive approach is proposed that combines advanced feature selection techniques with powerful neural network (NN) architectures. The selected attributes from this process are then transmitted to the succeeding tiers of a hybrid NN, which integrates convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM) components. The Analysis of Variance (ANOVA) + CNN-BiLSTM hybrid model demonstrates superior performance, producing exceptional results with a precision of 98.84%, accuracy of 99.25%, F1 score of 99.26%, and recall of 99.69%. The proposed hybrid model for detecting GPS spoofing attacks exhibits significant improvements in terms of prediction accuracy, true positive and false positive rates, as well as F1 score and recall values.

Channel estimation poses a main challenge in intelligent reflecting surface (IRS)-assisted millimeter wave (mmWave) multiuser multiple-input multiple-output (MIMO) systems due to the substantial number of antennas at the base station (BS) and the passive reflective elements within the IRS lacking sufficient signal processing capabilities. This article addresses this challenge by proposing a channel estimation technique for IRS-assisted mmWave MIMO systems. The problem of channel estimation is normally taken as a compressed sensing (CS) problem, typically addressed through algorithms such as Orthogonal Matching Pursuit (OMP), Generalized Approximate Message Passing (GAMP), and Vector Approximate Message Passing with Expectation-Maximization (EM-VAMP). EM-VAMP demonstrates better performance only when a Gaussian mixture (GM) distribution is chosen as the prior for the sparse channel, especially at high signal-to-noise ratios (SNRs). To address this, the article introduces the application of generalized linear models (GLMs), extensions of standard linear models, providing increased flexibility in modeling data that deviates from Gaussian distribution. Numerical results unveil that the proposed Its EM-VAMP-GLM is much more robust to the existing OMP,

The rapid increase in the number of cyber attacks and the emergence of various attack variations pose significant threats to the security of computer systems and networks. Various intrusion detection systems (IDS) are developed to defend computer systems and networks in response to these threats. One type of IDS, known as a host-based intrusion detection system (HIDS), focuses on securing a single host. Numerous HIDS have been proposed in the literature, incorporating various detection methods. This study develops multiple machine learning (ML) models and stacking ensemble based HIDS that can be used as detection methods in HIDS. Initially, n-grams, standard bag-of-words (BoW), binary BoW, probability BoW, and term frequency-inverse document frequency (TF-IDF) BoW methods are applied to the ADFA-LD and ADFA-WD datasets. Mutual information and k-means methods are used together for feature selection on the resulting BoW datasets. Individual models are created using either selected features or all features. Subsequently, the outputs of these individual models are used in extreme gradient boosting (XGBoost) and adaptive boosting (AdaBoost) models to develop stacking ensemble based models. The experimental results show that the best accuracy (ACC) among models using ADFA-LD based BoW datasets is achieved by the stacking ensemble based XGBoost model, which has an ACC of 0.9747. This XGBoost model utilizes the standard BoW dataset and selected features. Among models using ADFA-WD based BoW datasets, the stacking ensemble based XGBoost is also the most successful in terms of ACC, with an ACC of 0.9163, using the standard BoW dataset and all features.

Blockchain technology has produced effective solutions and provides security by using cryptographic tools for various applications, attracting attention from the academic community. Therefore, researchers have taken advantage of the features of blockchain technology to increase the security of the ecosystem. Recently, as the existence of quantum computers has been felt, researchers have started to benefit from post-quantum cryptography to increase privacy and security. There has been an increase in data and asset protection in post-quantum blockchain-based solutions. To the best of our knowledge, there is no comprehensive review or taxonomy that provides a complete picture of post-quantum secure structures with privacy-preserving techniques that have the potential to be used in blockchain. This paper aims to close this gap by systematically examining these approaches and revealing the deficiencies in the existing literature and the development potential in these areas. The taxonomy examines the role of blockchain technology in post-quantum cryptography and emphasizes the potential of technologies such as zero-knowledge proof to ensure privacy in post-quantum blockchain-based systems. We also review the existing literature on addressing the performance overhead, interoperability, scalability, and security challenges in implementing post-quantum cryptography in zero-knowledge proof-enabled blockchain architectures that protect against quantum computing threats. The studies are collected from journal papers in widely used academic databases between 2018 and 2024. The studies are subjected to certain elimination criteria, and 13 studies are reviewed in detail. Our approach will facilitate discussions on future research directions by proposing the accessibility of post-quantum cryptography against quantum threats to blockchain systems and solutions to the challenges that arise in the integration phase.

This paper proposes a quantum-secure, privacy-preserving blockchain platform for the Internet of Medical Things (IoMT). It defines a solution to quantum attacks on blockchain by integrating the Dilithium lattice-based signature scheme to enhance security and privacy. A layer-based structure, combined with the Constrained Application Protocol (CoAP), is used to improve the efficiency of data sharing, optimize security, and manage authentication in resource-constrained IoMT environments. Zero-knowledge proofs (ZKP) and lattice-based signatures are used for lightweight authentication and data integrity. Real-time testing on electrochemical sensor data validates the system’s efficiency in securely managing IoMT communications. Additionally, event-based smart contracts (EBSC) are implemented to reduce communication costs and minimize blockchain overhead. Experimental results show that Dilithium outperforms other schemes like Falcon and ECDSA, making it a superior solution for real-time IoMT security.

The Internet of Vehicles (IoV) is a specialized iteration of the Internet of Things (IoT) tailored to facilitate communication and connectivity among vehicles and their environment. It harnesses the power of advanced technologies such as cloud computing, wireless communication, and data analytics to seamlessly exchange real-time data among vehicles, road-side infrastructure, traffic management systems, and other entities. The primary objectives of this real-time data exchange include enhancing road safety, reducing traffic congestion, boosting traffic flow efficiency, and enriching the driving experience. Through the IoV, vehicles can share information about traffic conditions, weather forecasts, road hazards, and other relevant data, fostering smarter, safer, and more efficient transportation networks. Developing, implementing and maintaining sophisticated techniques for detecting attacks present significant challenges and costs, which might limit their deployment, especially in smaller settings or those with constrained resources. To overcome these drawbacks, this article outlines developing an innovative attack detection model for the IoV using advanced deep learning techniques. The model aims to enhance security in vehicular networks by efficiently identifying attacks. Initially, data is collected from online databases and subjected to an optimal feature extraction process. During this phase, the Enhanced Exploitation in Hybrid Leader-based Optimization (EEHLO) method is employed to select the optimal features. These features are utilized by a Complex-Valued Dilated Recurrent Neural Network (CV-DRNN) to detect attacks within vehicle networks accurately. The performance of this novel attack detection model is rigorously evaluated and compared with that of traditional models using a variety of metrics.