An effective technique for detecting minority attacks in NIDS using deep learning and sampling approach

Alexandria Engineering Journal, Год журнала: 2023, Номер 78, С. 469 - 482

Опубликована: Авг. 3, 2023

Anomaly-based intrusion detection system have been consistently used in business organizations and military to detect a breach network by identifying any activity that deviates from the baseline pattern. In this paper, we propose an effective technique identify predict minority attacks with three layers. Here, first layer utilizes Weighted Deep Neural Network (WDNN) for suspicious traffic samples it is passed second layer. Layer 2 classifies as normal or majority using Convolutional (CNN) Long-Short Term Memory (LSTM). Any sample classified attack sent 3 XGBoost algorithm. into their respective classes. To boost rate of attacks, employs One-Sided Selection under-sampling algorithm remove noisy An Adaptive Synthetic (ADASYN) oversampling generates synthetic evaluate system, datasets namely NSL KDD, CICIDS-2017 CIDDS 001 dataset are used. The attained overall accuracy 97.94% on KDD dataset, 98.3% 97.9% dataset.

Язык: Английский

---

Optimized MLP-CNN Model to Enhance Detecting DDoS Attacks in SDN Environment

Network, Год журнала: 2023, Номер 3(4), С. 538 - 562

Опубликована: Дек. 1, 2023

In the contemporary landscape, Distributed Denial of Service (DDoS) attacks have emerged as an exceedingly pernicious threat, particularly in context network management centered around technologies like Software-Defined Networking (SDN). With increasing intricacy and sophistication DDoS attacks, need for effective countermeasures has led to adoption Machine Learning (ML) techniques. Nevertheless, despite substantial advancements this field, challenges persist, adversely affecting accuracy ML-based DDoS-detection systems. This article introduces a model designed detect attacks. leverages combination Multilayer Perceptron (MLP) Convolutional Neural Network (CNN) enhance performance systems within SDN environments. We propose utilizing SHapley Additive exPlanations (SHAP) feature-selection technique employing Bayesian optimizer hyperparameter tuning optimize our model. To further solidify relevance approach environments, we evaluate by using open-source dataset known InSDN. Furthermore, apply CICDDoS-2019 dataset. Our experimental results highlight remarkable overall 99.95% with impressive 99.98% InSDN These outcomes underscore effectiveness proposed environments compared existing

Язык: Английский

---

Performance Evaluation of Deep Learning Models for Classifying Cybersecurity Attacks in IoT Networks

Informatics, Год журнала: 2024, Номер 11(2), С. 32 - 32

Опубликована: Май 17, 2024

The Internet of Things (IoT) presents great potential in various fields such as home automation, healthcare, and industry, among others, but its infrastructure, the use open source code, lack software updates make it vulnerable to cyberattacks that can compromise access data services, thus making an attractive target for hackers. complexity has increased, posing a greater threat public private organizations. This study evaluated performance deep learning models classifying cybersecurity attacks IoT networks, using CICIoT2023 dataset. Three architectures based on DNN, LSTM, CNN were compared, highlighting their differences layers activation functions. results show architecture outperformed others accuracy computational efficiency, with rate 99.10% multiclass classification 99.40% binary classification. importance standardization proper hyperparameter selection is emphasized. These demonstrate CNN-based model emerges promising option detecting cyber threats environments, supporting relevance network security.

Язык: Английский

---

A Systematic Review on Game-Theoretic Models and Different Types of Security Requirements in Cloud Environment: Challenges and Opportunities

Archives of Computational Methods in Engineering, Год журнала: 2024, Номер unknown

Опубликована: Апрель 1, 2024

Язык: Английский

---

Cloud Network Anomaly Detection Using Machine and Deep Learning Techniques— Recent Research Advancements

IEEE Access, Год журнала: 2024, Номер 12, С. 56749 - 56773

Опубликована: Янв. 1, 2024

In the rapidly evolving landscape of computing and networking, concepts cloud networks have gained significant prominence. Although network offers on-demand access to shared resources, anomalies pose potential risks integrity security networks. However, protecting against remains a challenge. Unlike traditional detection techniques, machine learning (ML) deep (DL) offer new adaptable methods for detecting in The objective this study is comprehensively explore existing ML /DL different based on distributed denial service anomaly (DDoS) intrusion systems (IDS) seeks address gaps networks, proposing solutions these environments. ultimate goal contribute valuable insights practical enhance reliability through effective by ML/ DL techniques. Methodologies ML/DL are explained, along with their advantages, disadvantages, respective approaches. addition, summary comparison between models also included.

Язык: Английский

---

Ensemble learning based anomaly detection for IoT cybersecurity via Bayesian hyperparameters sensitivity analysis

Cybersecurity, Год журнала: 2024, Номер 7(1)

Опубликована: Июнь 12, 2024

Abstract The Internet of Things (IoT) integrates more than billions intelligent devices over the globe with capability communicating other connected little to no human intervention. IoT enables data aggregation and analysis on a large scale improve life quality in many domains. In particular, collected by contain tremendous amount information for anomaly detection. heterogeneous nature is both challenge an opportunity cybersecurity. Traditional approaches cybersecurity monitoring often require different kinds pre-processing handling various types, which might be problematic datasets that features. However, types network can capture diverse set signals single type device readings, particularly useful this paper, we present comprehensive study using ensemble machine learning methods enhancing via Rather one model, combines predictive power from multiple models, their accuracy rather model. We propose unified framework utilises Bayesian hyperparameter optimisation adapt environment contains sensor readings. Experimentally, illustrate high when compared traditional methods.

Язык: Английский

---

Robust DDoS attack detection with adaptive transfer learning

Computers & Security, Год журнала: 2024, Номер 144, С. 103962 - 103962

Опубликована: Июнь 22, 2024

In the evolving cybersecurity landscape, rising frequency of Distributed Denial Service (DDoS) attacks requires robust defense mechanisms to safeguard network infrastructure availability and integrity. Deep Learning (DL) models have emerged as a promising approach for DDoS attack detection mitigation due their capability automatically learning feature representations distinguishing complex patterns within traffic data. However, effectiveness DL in protecting against depends also on design adaptive architectures, through combination appropriate models, quality data, thorough hyperparameter optimizations, which are scarcely performed literature. Also, architectures detection, no method has yet addressed how transfer knowledge between different datasets improve classification accuracy. this paper, we propose an innovative by leveraging Convolutional Neural Networks (CNN), techniques. Experimental results publicly available show that proposed effectively identifies benign malicious activities specific categories.

Язык: Английский

---

DIWGAN Optimized with Namib Beetle Optimization Algorithm for Intrusion Detection in Mobile Ad Hoc Networks

IETE Journal of Research, Год журнала: 2023, Номер 70(5), С. 4422 - 4441

Опубликована: Июль 12, 2023

Mobile ad hoc network (MANET) plays a major role in wireless devices such as defense and flooding. Despite their smart applications, MANET faces more security issues than traditional wired networks on account of distinct features, no central coordination, dynamic topology, temporal life, the nature communication. To overcome these issues, this manuscript proposes Dual Interactive Wasserstein Generative Adversarial Network optimized with Namib Beetle Optimization Algorithm is proposed for intrusion detection preventing attacks MANET. By utilizing One Way Hash Chain Function, mobile users first register Trusted Authority. Each user sends finger vein biometric along id, latitude, longitude authentication verification. The packet analyzer, feature extraction, preprocessing, classification are four parts that make up detection. determine if any attack patterns have been identified, analyzer examined. This executed using Type 2 Fuzzy Controller deems header information. Anisotropic diffusion Kuwahara filtering techniques time series taken into consideration preprocessing unit. battle royal optimization algorithm utilized extraction unit to acquire better collection features categorization. classifies packets five categories: DoS, Probe, U2R, R2L, Anomaly technique. Finally, method provides 26.26%, 15.57%, 32.9% higher accuracy, 33.06%, 23.82%, 38.84% lesser delay analysed existing models.

Язык: Английский

---

An Unsupervised Approach for the Detection of Zero-Day DDoS Attacks in IoT Networks

Authorea (Authorea), Год журнала: 2024, Номер unknown

Опубликована: Янв. 14, 2024

In this article, an unsupervised IDS (Intrusion Detection System) is presented for the detection of zero-day DDoS (Distributed Denial Service) attacks IoT (Internet Things) networks that can detect anomalies without need prior knowledge or training in attack information. Attackers exploit existing undiscovered vulnerabilities system to launch attacks. There exist many traditional deep learning and machine based systems cannot deal with new mostly misclassify those Zero-day are often unknown threats have not been encountered before, addition, labelling data a time-consuming task security experts, So there exists methods unseen cyber-attacks on zero-day. recently adversely affected organisations terms finance services, as these become more sophisticated damaging. The growth has facilitated work, approach-based algorithm proposed by exploiting random projection feature selection process reduce dimensionality network ensemble model consisting K-means, GMM one-class SVM classification normal using hard voting technique. CIC-DDoS2019 datasets used extensive evaluation method. method obtained accuracy 94.55%, which better than other state-of-the-art learning-based methods.

Язык: Английский

---

Strengthening Network Security: Deep Learning Models for Intrusion Detection with Optimized Feature Subset and Effective Imbalance Handling

Computers, materials & continua/Computers, materials & continua (Print), Год журнала: 2024, Номер 78(2), С. 1995 - 2022

Опубликована: Янв. 1, 2024

In recent years, frequent network attacks have highlighted the importance of efficient detection methods for ensuring cyberspace security.This paper presents a novel intrusion system consisting data preprocessing stage and deep learning model accurately identifying attacks.We proposed four neural models, which are constructed using architectures such as Convolutional Neural Networks (CNN), Bi-directional Long Short-Term Memory (BiLSTM), Bidirectional Gate Recurrent Unit (BiGRU), Attention mechanism.These models been evaluated their performance on NSL-KDD dataset.To enhance compatibility between we apply various techniques employ particle swarm optimization algorithm to perform feature selection dataset, resulting in an optimized subset.Moreover, address class imbalance dataset focal loss.Finally, BO-TPE optimize hyperparameters maximizing performance.The test results demonstrate that is capable extracting spatiotemporal features traffic effectively.In binary multiclass experiments, it achieved accuracy rates 0.999158 0.999091, respectively, surpassing other state-of-the-art methods.

Язык: Английский

---