A deep learning-based approach with two-step minority classes prediction for intrusion detection in Internet of Things networks

Knowledge-Based Systems, Год журнала: 2025, Номер unknown, С. 113143 - 113143

Опубликована: Фев. 1, 2025

Язык: Английский

---

Sql injection detection algorithm based on Bi-LSTM and integrated feature selection

The Journal of Supercomputing, Год журнала: 2025, Номер 81(4)

Опубликована: Март 12, 2025

Язык: Английский

---

NSGTO‐LSTM: Niche‐strategy‐based gorilla troops optimization and long short‐term memory network intrusion detection model

ETRI Journal, Год журнала: 2025, Номер unknown

Опубликована: Март 17, 2025

Abstract In recent decades, the rapid growth of Internet Things (IoT) has highlighted several network security problems. this study, an efficient intrusion detection (ID) system is implemented by using both machine learning and data mining concepts for detecting patterns. During initial phase, are collected from NSL‐KDD University New South Wales‐Network Based 15 (UNSW‐NB15) datasets. The then normalized/scaled employing a standard scaler technique. Next, informative feature values selected proposed optimization algorithm—that is, Niche‐Strategy‐based Gorilla Troops Optimization (NSGTO) algorithm. Finally, these transferred to Long Short‐Term Memory (LSTM) model classify types attacks on comparison existing ID systems, based NSGTO‐LSTM obtains classification accuracy 99.98% 99.90%

Язык: Английский

---

Advancements in Machine Learning-Based Intrusion Detection in IoT: Research Trends and Challenges

Algorithms, Год журнала: 2025, Номер 18(4), С. 209 - 209

Опубликована: Апрель 9, 2025

This paper presents a systematic literature review based on the PRISMA model machine learning-based Distributed Denial of Service (DDoS) attacks in Internet Things (IoT) networks. The primary objective is to compare research trends deployment options, datasets, and learning techniques used domain between 2019 2024. results highlight dominance certain datasets (BoT-IoT TON_IoT) combination with Decision Tree (DT) Random Forest (RF) models, achieving high median accuracy rates (>99%). discusses various that are train evaluate (ML) models for detecting networks how they impact performance. Furthermore, findings suggest due hardware limitations, there preference lightweight ML solutions preprocessed datasets. Current indicate larger or industry-specific will continue gain popularity alongside more complex such as deep learning. emphasizes need robust scalable Software-Defined Networks (SDNs) offering flexibility, edge computing being extensively explored cloud environments, blockchain-integrated emerging promising approach enhancing security.

Язык: Английский

---

An Adaptive Framework for Intrusion Detection in IoT Security Using MAML (Model-Agnostic Meta-Learning)

Sensors, Год журнала: 2025, Номер 25(8), С. 2487 - 2487

Опубликована: Апрель 15, 2025

With the rapid emergence of Internet Things (IoT) devices, there were new vectors for attacking cyber, so was a need approachable intrusion detection systems (IDSs) with more innovative custom tactics. The traditional IDS models tend to find difficulties in generalization continuously changing and heterogeneous IoT environments. This paper contributes an adaptive framework using Model-Agnostic Meta-Learning (MAML) few-shot learning paradigms quickly adapt tasks little data. goal this research is improve security by developing strong that will perform well across assorted datasets attack Finally, we apply our proposed two benchmark datasets, UNSW-NB15 NSL-KDD99, which provide different scenarios network behaviors. methodology trains base model MAML allow fast adaptation on specific during fine-tuning. Our approach leads experimental results 99.98% accuracy, 99.5% precision, 99.0% recall, 99.4% F1 score dataset. achieved 99.1% 97.3% 98.2% 98.5% NSL-KDD99 That shows can detect many cyber threats Based study, it concluded meta-learning-based could help build resilient systems. Future works move educated meta-learning federated setting deploy real time response threats.

Язык: Английский

---

AE-MCDM: an autoencoder-based multi-criteria decision-making approach for unsupervised feature selection

The Journal of Supercomputing, Год журнала: 2025, Номер 81(7)

Опубликована: Май 1, 2025

Язык: Английский

---

RF‐FLIDS: A Novel Hybrid Intrusion Detection Model for Enhanced Anomaly Detection in IoT Networks

Security and Privacy, Год журнала: 2025, Номер 8(3)

Опубликована: Май 1, 2025

ABSTRACT The rapid proliferation of Internet Things (IoT) technology has transformed modern industries, introducing significant security vulnerabilities that necessitate practical intrusion detection systems (IDSs). While random forest (RF) shown promise for anomaly in IoT networks, its inherent bias toward majority classes imbalanced datasets limits effectiveness. Existing solutions primarily focus on data‐level techniques, such as oversampling or undersampling, algorithm‐level adjustments, including cost‐sensitive learning, which often increase time risk overfitting. This paper proposes RF‐FLIDS, a hybrid and Fuzzy Logic‐based IDS, to address this challenge, providing model‐level correction class imbalance. RF‐FLIDS mitigates the overfitting risks associated with eliminates need reweighting modifying RF algorithm, making it well suited real‐time detection. model was evaluated using two benchmark datasets, CICIoT2023 Aposemat IoT‐23, compared against traditional machine learning models, Balanced Random Forest (BRF), logistic regression (LR), k‐nearest neighbor (KNN), decision trees (DT), Gaussian Naive Bayes (GNB). Experimental results demonstrated superior performance achieving Accuracy 92.52% 92.92%, improvements Precision, Recall, F1‐score. These findings confirm robustness adaptability proposed enhancing security.

Язык: Английский

---

Enhancing IoT Security Using GA-HDLAD: A Hybrid Deep Learning Approach for Anomaly Detection

Applied Sciences, Год журнала: 2024, Номер 14(21), С. 9848 - 9848

Опубликована: Окт. 28, 2024

The adoption and use of the Internet Things (IoT) have increased rapidly over recent years, cyber threats in IoT devices also become more common. Thus, development a system that can effectively identify malicious attacks reduce security has topic great importance. One most serious comes from botnets, which commonly attack by interrupting networks required for to run. There are number methods be used improve identifying unknown patterns networks, including deep learning machine approaches. In this study, an algorithm named genetic with hybrid learning-based anomaly detection (GA-HDLAD) is developed, aim improving botnets within environment. GA-HDLAD technique addresses problem high dimensionality using during feature selection. Hybrid detect botnets; approach combination recurrent neural (RNNs), extraction techniques (FETs), attention concepts. Botnet involve complex (HDL) method detect. Moreover, FETs model ensures features extracted spatial data, while temporal dependencies captured RNNs. Simulated annealing (SA) utilized select hyperparameters necessary HDL approach. experimentally assessed benchmark botnet dataset, findings reveal provides superior results comparison existing methods.

Язык: Английский

---

A comprehensive analysis of machine learning-based intrusion detection systems: evaluating datasets and algorithms for internet of things

Journal of Cyber Security Technology, Год журнала: 2024, Номер unknown, С. 1 - 27

Опубликована: Дек. 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.

Язык: Английский

---

A hybrid machine learning approach for feature selection in designing intrusion detection systems (IDS) model for distributed computing networks

The Journal of Supercomputing, Год журнала: 2024, Номер 81(1)

Опубликована: Дек. 7, 2024

Язык: Английский

---