Explainable Artificial Intelligence in CyberSecurity: A Survey

IEEE Access, Год журнала: 2022, Номер 10, С. 93575 - 93600

Опубликована: Янв. 1, 2022

Nowadays, Artificial Intelligence (AI) is widely applied in every area of human being's daily life. Despite the AI benefits, its application suffer from opacity complex internal mechanisms and doesn't satisfy by design principles Explainable (XAI). The lack transparency further exacerbates problem field Cybersecurity because entrusting crucial decisions to a system that cannot explain itself presents obvious dangers. There are several methods literature capable providing explainability results. Anyway, XAI can be double-edged sword. It substantially improves practices but simultaneously leaves vulnerable adversary attacks. Therefore, there need analyze state-of-the-art provide clear vision for future research. This study an in-depth examination Cybersecurity. considers more than 300 papers comprehensively main fields, like Intrusion Detection Systems,Malware detection, Phishing Spam BotNets Fraud Zero-Day vulnerabilities, Digital Forensics Crypto-Jacking. Specifically, this focuses on adopted or proposed these pointing out promising works new challenges.

Язык: Английский

---

Detecting Vulnerability on IoT Device Firmware: A Survey

IEEE/CAA Journal of Automatica Sinica, Год журнала: 2022, Номер 10(1), С. 25 - 41

Опубликована: Сен. 6, 2022

Internet of things (IoT) devices make up 30% all network-connected endpoints, introducing vulnerabilities and novel attacks that many companies as primary targets for cybercriminals. To address this increasing threat surface, every organization deploying IoT needs to consider security risks ensure those are secure trusted. Among the solutions risks, firmware analysis is essential fix software bugs, patch vulnerabilities, or add new features protect users vulnerable devices. However, has never been an easy job due diversity execution environment close source firmware. These two distinct complicate operations unpack samples detailed analysis. They also it difficult create visual environments emulate running device Although researchers have developed methods overcome various challenges in past decade, critical barriers impede practice. Therefore, survey motivated systematically review analyze research their solutions, considering both breadth depth. Specifically, based on perspectives, perform introduced classified into four categories. The each category discussed detail, potential proposed subsequently. We then discuss flaws these provide future directions field. This can be utilized by a broad range readers, including developers, cyber researchers, engineers, better understand

Язык: Английский

---

A Review on Cybersecurity Analysis, Attack Detection, and Attack Defense Methods in Cyber-physical Power Systems

Journal of Modern Power Systems and Clean Energy, Год журнала: 2023, Номер 11(3), С. 727 - 743

Опубликована: Янв. 1, 2023

Potential malicious cyber-attacks to power systems which are connected a wide range of stakeholders from the top tail will impose significant societal risks and challenges. The timely detection defense crucial importance for safe reliable operation cyber-physical (CPPSs). This paper presents comprehensive review some latest attack strategies. Firstly, vulnerabilities brought by new information communication technologies (ICTs) analyzed, their impacts on security CPPSs discussed. Various cyber physical layers then analyzed within framework, features negative Secondly, two current mainstream methods including state estimation based machine learning benefits drawbacks Moreover, active passive comprehensively Finally, trends challenges in strategies provided.

Язык: Английский

---

A Review of Anomaly Detection Strategies to Detect Threats to Cyber-Physical Systems

Electronics, Год журнала: 2023, Номер 12(15), С. 3283 - 3283

Опубликована: Июль 30, 2023

Cyber-Physical Systems (CPS) are integrated systems that combine software and physical components. CPS has experienced rapid growth over the past decade in fields as disparate telemedicine, smart manufacturing, autonomous vehicles, Internet of Things, industrial control systems, power grids, remote laboratory environments, many more. With widespread integration various aspects contemporary society, frequency malicious assaults carried out by adversaries a substantial surge recent times. Incidents targeting vital civilian infrastructure, such electrical grids oil pipelines, have become alarmingly common due to expanded connectivity public internet, which significantly expands vulnerability CPS. This article presents comprehensive review existing literature examines latest advancements anomaly detection techniques for identifying security threats Systems. The primary emphasis is placed on addressing life safety concerns within networks (ICS). A total 296 papers reviewed, with themes research gaps identified. paper makes novel contribution key challenges remain field, include resource constraints, lack standardized communication protocols, extreme heterogeneity hampers industry consensus, different information priorities between Operational Technology (OT) Information (IT) networks. Potential solutions and/or opportunities further identified address these selected challenges.

Язык: Английский

---

Graph Neural Networks for Intrusion Detection: A Survey

IEEE Access, Год журнала: 2023, Номер 11, С. 49114 - 49139

Опубликована: Янв. 1, 2023

Cyberattacks represent an ever-growing threat that has become a real priority for most organizations. Attackers use sophisticated attack scenarios to deceive defense systems in order access private data or cause harm. Machine Learning (ML) and Deep (DL) have demonstrate impressive results detecting cyberattacks due their ability learn generalizable patterns from flat data. However, fail capture the structural behavior of attacks, which is essential effective detection. Contrarily, graph structures provide more robust abstract view system difficult attackers evade. Recently, Graph Neural Networks (GNNs) successful learning useful representations semantic provided by graph-structured Intrusions been detected years using graphs such as network flow provenance graphs, these can help models understand addition traditional features. In this survey, we focus on applications representation detection network-based host-based intrusions, with special attention GNN methods. For both host levels, present be leveraged comprehensively review state-of-the-art papers along used datasets. Our analysis reveals GNNs are particularly efficient cybersecurity, since they without requiring any external domain knowledge. We also evaluate robustness techniques based adversarial attacks. Finally, discuss strengths weaknesses GNN-based intrusion identify future research directions.

Язык: Английский

---

Current trends in AI and ML for cybersecurity: A state-of-the-art survey

Cogent Engineering, Год журнала: 2023, Номер 10(2)

Опубликована: Окт. 25, 2023

Язык: Английский

---

Security of networked control systems subject to deception attacks: a survey

International Journal of Systems Science, Год журнала: 2022, Номер 53(16), С. 3577 - 3598

Опубликована: Ноя. 12, 2022

A networked control system (NCS), which integrates various physical components by utilising communication networks, is a complex intelligent with high flexibility and reliability. It has been widely applied in areas, such as power grids, transportation, smart manufacturing. However, compared traditional systems, NCSs expose more extra vulnerabilities the openness of leading to long-term concerns for security against cyber attacks. This paper gives survey detail on recent developments subject deception attacks from two domains information technology (IT) control, respectively. First, several incidents reported years are reviewed couple prevailing analysed. Besides, results IT respect protection-detection-reaction model summarised. Then, domain issues attack design, detection, secure state estimation resilient surveyed depth. Furthermore, novel topics combination also discussed. Finally, future research directions presented this topic.

Язык: Английский

---

Fixed-Time Resilient Edge-Triggered Estimation and Control of Surface Vehicles for Cooperative Target Tracking Under Attacks

IEEE Transactions on Intelligent Vehicles, Год журнала: 2022, Номер 8(1), С. 547 - 556

Опубликована: Июнь 17, 2022

This paper is concerned with the cooperative target tracking of under-actuated unmanned surface vehicles (USVs) event-triggered communications subject to denial-of-service (DoS) attacks. The position information can be sensed by a fraction follower USVs only. A fixed-time resilient edge-triggered estimation and control architecture presented for achieving under DoS Specifically, distributed extended state observer (ESO) designed recover velocity prescribed time regardless unreliable communication network Moreover, burden reduced proposed mechanism. In law design, ESO estimating model uncertainties external disturbances in an earth-fixed reference frame. Then, each USV based on ESO. It proven that error signals closed-loop system are convergent origin fixed time. An example provided substantiate effectiveness USVs.

Язык: Английский

---

A General Safety-Certified Cooperative Control Architecture for Interconnected Intelligent Surface Vehicles With Applications to Vessel Train

IEEE Transactions on Intelligent Vehicles, Год журнала: 2022, Номер 7(3), С. 627 - 637

Опубликована: Апрель 22, 2022

This paper considers cooperative control of interconnected intelligent surface vehicles (ISV) moving in a complex water containing multiple static/dynamic obstacles. Each ISV is subject to force and moment constraints, addition internal model uncertainties external disturbances induced by wind, waves currents. A general safety-certified architecture capable achieving various collective behaviors such as consensus, containment, enclosing, flocking, proposed. Specifically, distributed motion generator used generate desired reference signals for each ISV. Robust-exact-differentiators-based (RED-based) extended state observers (ESOs) are designed recovering unknown total finite time. With the aid Lyapunov functions, input-to-state safe high order barrier functions RED-based ESOs, constrained quadratic optimization problems formulated optimal surge yaw without violating input, stability, safety constraints. In facilitate real-time implementations, one-layer recurrent neural network employed solve problem on board. It proved that all tracking errors closed-loop system uniformly ultimately bounded multi-ISV safe. An example given substantiate effectiveness proposed architecture.

Язык: Английский

---

Intrusion Detection in IoT Using Deep Learning

Sensors, Год журнала: 2022, Номер 22(21), С. 8417 - 8417

Опубликована: Ноя. 2, 2022

Cybersecurity has been widely used in various applications, such as intelligent industrial systems, homes, personal devices, and cars, led to innovative developments that continue face challenges solving problems related security methods for IoT devices. Effective methods, deep learning intrusion detection, have introduced. Recent research focused on improving algorithms improved IoT. This explores detection implemented using learning, compares the performance of different identifies best method implementing is conducted models based convolutional neural networks (CNNs), long short-term memory (LSTM), gated recurrent units (GRUs). A standard dataset considered evaluate proposed model. Finally, empirical results are analyzed compared with existing approaches The seemed highest accuracy methods.

Язык: Английский

---