A comprehensive analysis of machine learning-based intrusion detection systems: evaluating datasets and algorithms for internet of things DOI
Sohail Saif, Amir H. Ansari, Suparna Biswas

et al.

Journal of Cyber Security Technology, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 27

Published: Dec. 27, 2024

With the recent advancement of Internet Things (IoT) in various sectors, security has become an essential requirement. Any IoT application or device may be compromised by intruders to disrupt entire network. These kinds insider attacks are difficult prevent. Here, Intrusion Detection System (IDS) can play important role identifying unknown attacks. IDS uses network traffic logs detect and respond suspicious activities anomalies before attackers exploit system weaknesses. Machine learning models among most efficient effective methods identify anomalous behaviors. Hence, this paper, we have conducted a comprehensive analysis utilizing several supervised semi-supervised machine algorithms assess their performance. We utilized 15 benchmark datasets containing samples related employed holdout k-fold cross-validation for performance comparison. also discussed identified possible reasons respective outcomes. Experimental results indicate that two algorithms, kNN ANN, exhibit highest terms accuracy, precision, recall, etc. This with evaluation metrics provides researchers valuable insights.

Language: Английский

A hybrid approach for efficient feature selection in anomaly intrusion detection for IoT networks DOI Creative Commons
Aya G. Ayad, Nehal A. Sakr, Noha A. Hikal

et al.

The Journal of Supercomputing, Journal Year: 2024, Volume and Issue: 80(19), P. 26942 - 26984

Published: Aug. 29, 2024

Abstract The exponential growth of Internet Things (IoT) devices underscores the need for robust security measures against cyber-attacks. Extensive research in IoT community has centered on effective traffic detection models, with a particular focus anomaly intrusion systems (AIDS). This paper specifically addresses preprocessing stage datasets and feature selection approaches to reduce complexity data. goal is develop an efficient AIDS that strikes balance between high accuracy low time. To achieve this goal, we propose hybrid approach combines filter wrapper methods. integrated into two-level system. At level 1, our classifies network packets normal or attack, 2 further classifying attack determine its specific category. One critical aspect consider imbalance these datasets, which addressed using Synthetic Minority Over-sampling Technique (SMOTE). evaluate how selected features affect performance machine learning model across different algorithms, namely Decision Tree, Random Forest, Gaussian Naive Bayes, k-Nearest Neighbor, employ benchmark datasets: BoT-IoT, TON-IoT, CIC-DDoS2019. Evaluation metrics encompass accuracy, precision, recall, F1-score. Results indicate decision tree achieves ranging 99.82 100%, short times 0.02 0.15 s, outperforming existing architectures networks establishing superiority achieving both times.

Language: Английский

Citations

7
A Multiscale Principal Component Analysis Approach to Physical Layer Secret Key Generation in Indoor Environments DOI Open Access
Megha S. Kumar,

R. Ramanathan

Transactions on Emerging Telecommunications Technologies, Journal Year: 2025, Volume and Issue: 36(3)

Published: March 1, 2025

ABSTRACT With the rise of Industry 5.0, smart cities, and ever‐expanding use general wireless networks, ensuring seamless communication robust data security has become a critical challenge. Generating secure secret keys (SKG) through channels is particularly complex in environments where noise wideband conditions introduce discrepancies autocorrelation channel measurements. These issues compromise cross‐correlation randomness, leading to substantial bit disagreements, distinct at transceivers, unsuccessful SKG. This research begins by outlining mathematical model signal preprocessing technique called multiscale principal component analysis (MSPCA). Subsequently, it explores performance key generation when employing proposed scheme. A holistic system‐level framework for creating initial shared presented, encompassing quantization methods such as uniform multilevel (UMQ) encoding 3‐bit Gray encoding. Monte Carlo‐based simulations an indoor scenario evaluate system efficacy using metrics like Pearson correlation coefficient, disagreement rate (BDR), complexity. The scheme achieves BDR lower than 0.01, coefficient greater 0.95, passes all National Institute Standards Technology (NIST) randomness tests, establishing viable solution securing systems. In context 5.0 city infrastructures, are paramount, SKG offers significant potential. its ability ensure reliable communication, this can underpin development advanced systems that cater high demands interconnected ecosystems, enhancing resilience trust applications.

Language: Английский

Citations

0
A feature selection-driven machine learning framework for anomaly-based intrusion detection systems DOI Creative Commons

Emre Emirmahmutoğlu,

Yılmaz Atay

Peer-to-Peer Networking and Applications, Journal Year: 2025, Volume and Issue: 18(3)

Published: April 28, 2025

Language: Английский

Citations

0
A Hybrid Feature Selection Model for Anomaly-Based Intrusion Detection in IoT Networks DOI
Aya G. Ayad, Nehal A. Sakr, Noha A. Hikal

et al.

2022 International Telecommunications Conference (ITC-Egypt), Journal Year: 2024, Volume and Issue: unknown, P. 1 - 7

Published: July 22, 2024

Language: Английский

Citations

2
A hybrid intrusion detection approach based on message queuing telemetry transport (MQTT) protocol in industrial internet of things DOI

Georg Thamer Francis,

Alireza Souri, Nihat İnanç

et al.

Transactions on Emerging Telecommunications Technologies, Journal Year: 2024, Volume and Issue: 35(9)

Published: Aug. 20, 2024

Abstract The number of attacks against Industrial Internet Things (IIoT) devices has increased over the past years, particularly on widely used communication protocols like Message Queuing Telemetry Transfer (MQTT). fast increase in IIoT applications brings both critical challenges and technical gaps cybersecurity. On other hand, traditional cyber‐attack detection approaches scrap to address support run‐time responsibilities environments. This study presents a hybrid Genetic Algorithm Random Forest (GA_RF) method for detecting cyber‐attacks Control Machines (ICS) that use MQTT protocol environment. architecture integrates ICS with edge cloud servers, using GA_RF algorithm detect anomalies data collected by sensors. Normal is processed locally then sent storage return, ensuring continuous monitoring security. Also, MQTT‐IOT‐IDS2020 dataset as real test case was applied prediction proposed compare some powerful machine deep learning models. experimental results show an optimum accuracy 99.87%–100% cyber‐attacks. also achieved 0–0.0015 Mean Absolute Error (MAE) 100% Precision, Recall, F‐score factors. result led architecture, which connects server while running In conclusion, this indicates effectiveness aims improve security IIoT.

Language: Английский

Citations

1
MS-CFFS: Multistage Coarse and Fine Feature Selecton for Advanced Anomaly Detection in IoT Security Networks DOI Creative Commons
Mohammed Sayeeduddin Habeeb,

Tummala Ranga Babu

International Journal of Electrical and Electronics Research, Journal Year: 2024, Volume and Issue: 12(3), P. 780 - 790

Published: July 25, 2024

In recent years, the concept of Internet-of-Things (IoT) has increased in popularity, leading to a massive increase both number connected devices and volume data they handle. With IoT constantly collecting sharing large quantities sensitive data, securing this is major concern, especially with network anomalies. A network-based anomaly detection system serves as crucial safeguard for networks, aiming identify irregularities entry point by continuously monitoring traffic. However, research community contributed more field, security still faces several challenges detecting these anomalies, often resulting high rate false alarms missed detections when it comes classifying traffic computational complexity. Seeing this, we propose novel method capabilities Anomaly Detection IoT. This study introduces deep learning (DL) based Multistage Coarse Fine Feature Selection (MS-CFFS), improve techniques devised frameworks. The proposed feature section done two stages. MS-CFFS, utilizing learning-based dual-stage selection, substantially improves NIDS efficacy. results confirm MS-CFFS's outstanding classification accuracy at 99.93%, remarkably low FAR 0.05% FNR 0.11%. These achievements stem from refining set 28 pivotal features, thus notably cutting complexity without sacrificing precision. Furthermore, comparative analysis leading-edge approaches validates preeminence our MS-CFFS domain security.

Language: Английский

Citations

0
A comprehensive analysis of machine learning-based intrusion detection systems: evaluating datasets and algorithms for internet of things DOI
Sohail Saif, Amir H. Ansari, Suparna Biswas

et al.

Journal of Cyber Security Technology, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 27

Published: Dec. 27, 2024

With the recent advancement of Internet Things (IoT) in various sectors, security has become an essential requirement. Any IoT application or device may be compromised by intruders to disrupt entire network. These kinds insider attacks are difficult prevent. Here, Intrusion Detection System (IDS) can play important role identifying unknown attacks. IDS uses network traffic logs detect and respond suspicious activities anomalies before attackers exploit system weaknesses. Machine learning models among most efficient effective methods identify anomalous behaviors. Hence, this paper, we have conducted a comprehensive analysis utilizing several supervised semi-supervised machine algorithms assess their performance. We utilized 15 benchmark datasets containing samples related employed holdout k-fold cross-validation for performance comparison. also discussed identified possible reasons respective outcomes. Experimental results indicate that two algorithms, kNN ANN, exhibit highest terms accuracy, precision, recall, etc. This with evaluation metrics provides researchers valuable insights.

Language: Английский

Citations

0