Enhanced Hybrid Deep Learning Models-Based Anomaly Detection Method for Two-Stage Binary and Multi-Class Classification of Attacks in Intrusion Detection Systems

Algorithms, Год журнала: 2025, Номер 18(2), С. 69 - 69

Опубликована: Янв. 28, 2025

As security threats become more complex, the need for effective intrusion detection systems (IDSs) has grown. Traditional machine learning methods are limited by extensive feature engineering and data preprocessing. To overcome this, we propose two enhanced hybrid deep models, an autoencoder–convolutional neural network (Autoencoder–CNN) a transformer–deep (Transformer–DNN). The Autoencoder reshapes traffic data, addressing class imbalance, CNN performs precise classification. transformer component extracts contextual features, which DNN uses accurate Our approach utilizes adaptive synthetic sampling–synthetic minority oversampling technique (ADASYN-SMOTE) binary classification SMOTE multi-class classification, along with edited nearest neighbors (ENN) further imbalance handling. models were designed to minimize false positives negatives, improve real-time detection, identify zero-day attacks. Evaluations based on CICIDS2017 dataset showed 99.90% accuracy Autoencoder–CNN 99.92% Transformer–DNN in 99.95% 99.96% respectively. On NF-BoT-IoT-v2 dataset, achieved 99.98% 97.95% while reached 97.90%, These results demonstrate superior performance of proposed compared traditional handling diverse

Язык: Английский

---

Enhancing Multi-Class Attack Detection in Graph Neural Network through Feature Rearrangement

Electronics, Год журнала: 2024, Номер 13(12), С. 2404 - 2404

Опубликована: Июнь 19, 2024

As network sizes grow, attack schemes not only become more varied but also increase in complexity. This diversification leads to a proliferation of variants, complicating the identification and differentiation potential threats. Enhancing system security necessitates implementation multi-class intrusion detection systems. approach enables categorization incoming traffic into distinct types illustrates specific encountered within Internet. Numerous studies have leveraged deep learning (DL) for Network-based Intrusion Detection Systems (NIDS), aiming improve detection. Among these DL algorithms, Graph Neural Networks (GNN) stand out their ability efficiently process unstructured data, especially traffic, making them particularly suitable NIDS applications. Although usually monitors outgoing flows network, represented as edge features graph format, traditional GNN consider node features, overlooking features. oversight can result losing important flow data diminish system’s detect attacks effectively. To address this limitation, our research makes several key contributions: (1) Emphasize significance enhancing detection, (2) Utilize port information, which is essential identifying often overlooked during training, (3) Reorganize embedded graph. By doing this, represent close actual showing endpoint information such IP addresses ports; contains related Duration, Number Packet/s, Length…; (4) Compared methods, experiments demonstrate significant performance improvements on both CIC-IDS-2017 (98.32%) UNSW-NB15 (96.71%) datasets.

Язык: Английский

---

Comparative Analysis of Advanced Machine Learning Models for Exploit Detection in Intrusion Detection Systems

Research Square (Research Square), Год журнала: 2025, Номер unknown

Опубликована: Апрель 10, 2025

The integrity of network infrastructure against malicious exploit attacks relies mostly on Intrusion Detection Systems (IDS). These techniques are very essential for identifying and lowering threats before they start to cause significant damage. This manuscript evaluates three advanced Machine Learning (ML) models CatBoost, XGBoost, Long Short-Term Memory (LSTM) a real-world traffic dataset determine their suitability IDS applications. Every model is evaluated using key metrics: accuracy, precision, recall, F1-score, error measures including Root Mean Squared Error (RMSE) (MSE). Based the results, Catboost exceeds other with 98.55% accuracy lowest rates. Given CatBoost's remarkable performance, it fitting real-time systems where reducing false positives negatives extremely crucial. XGBoost provides balanced computationally affordable solution even if significantly less accurate; ideal scenarios requiring fast responses limited resources. Strong in sequential pattern recognition, LSTM has higher rate positives, suggesting that further tuning needed improve its overall reliability surroundings. possibility enhancing performance gradient boosting such as CatBoost cybersecurity underlined this study.

Язык: Английский

---

Classification of Multiclass DDOS Attack Detection Using Bayesian Weighted Random Forest Optimized With Gazelle Optimization Algorithm

Transactions on Emerging Telecommunications Technologies, Год журнала: 2025, Номер 36(4)

Опубликована: Апрель 1, 2025

ABSTRACT The increase in Distributed Denial of Service (DDoS) attacks poses a considerable threat to the security and stability current network, especially Internet Things (IoT) cloud environments. Traditional detection methods often struggle with inability achieve balance between accuracy computational efficiency. In this manuscript, Classification Multiclass DDOS Attack Detection using Bayesian Weighted Random Forest Optimized Gazelle Optimization Algorithm (DDOS‐AD‐BWRF‐GOA) is proposed. First, raw data gathered from CICDDoS2019 dataset. Then, input are preprocessed utilizing Adaptive Bitonic Filtering for normalizing values. fed Improved Feed Forward Long Short‐Term Memory technique selecting features that model's execution time. selected supplied (BWRF), which classifies multiclass attack. general, does not adopt any optimization define optimal parameters guarantee exact identification. Hence, GOA proposed optimize classifier. method implemented MATLAB. performance metrics, such as Accuracy, Precision, Recall, F 1‐score, Specificity, Error rate, Computational time evaluated. attains 15.34%, 24.1%, 18.9% higher 12.4%, 18.24%, 22.6% precision when analyzed existing techniques: Hybrid deep learning classification (HDL‐DDOS‐DC), Edge‐HetIoT Defense against DDoS attack techniques (EHD‐DDOS‐LT), Digital twin‐enabled intelligent autonomous core networks (DTI‐DDOS‐ACN), respectively.

Язык: Английский

---

Transformative role of private AI in secondary education

AIP conference proceedings, Год журнала: 2025, Номер 3157, С. 020004 - 020004

Опубликована: Янв. 1, 2025

Язык: Английский

---

Block chain role in real estate

AIP conference proceedings, Год журнала: 2025, Номер 3157, С. 100003 - 100003

Опубликована: Янв. 1, 2025

Язык: Английский

---

Boosting in high dimensional data classification

AIP conference proceedings, Год журнала: 2025, Номер 3157, С. 080008 - 080008

Опубликована: Янв. 1, 2025

---

Machine learning algorithms for personalized QoS aware web service recommendation

AIP conference proceedings, Год журнала: 2025, Номер 3157, С. 080012 - 080012

Опубликована: Янв. 1, 2025

Язык: Английский

---

Implementation of YOLO8 for real-time object recognition and tracking for visually impaired

AIP conference proceedings, Год журнала: 2025, Номер 3157, С. 070014 - 070014

Опубликована: Янв. 1, 2025

Язык: Английский

---

Cloud task scheduling using context-aware task scheduling with machine learning (CATSM-ML)

AIP conference proceedings, Год журнала: 2025, Номер 3157, С. 080009 - 080009

Опубликована: Янв. 1, 2025

Язык: Английский

---