Effective injection of adversarial botnet attacks in IoT ecosystem using evolutionary computing

Internet Technology Letters, Journal Year: 2023, Volume and Issue: 6(4)

Published: May 5, 2023

Abstract With the widespread adoption of Internet Things (IoT) technologies, botnet attacks have become most prevalent cyberattack. In order to combat attacks, there has been a considerable amount research on in IoT ecosystems by graph‐based machine learning (GML). The majority GML models are vulnerable adversarial (ADAs). These ADAs were created assess robustness existing ML‐based security solutions. this letter, we present novel attack (ADBA) that modifies graph data structure using genetic algorithms (GAs) trick detection system. According experiment results and comparative analysis, proposed ADBA can be executed resource‐constrained nodes. It offers substantial performance gain 2.15 s, 52 kb , 92 817 mJ 97.8%, 27.74%–41.82% over other approaches term Computing Time (CT), Memory Usage (MU), Energy (EU), Attack Success Rate (ASR) Accuracy (ACC) metrics, respectively.

Language: Английский

---

Securing ResNet50 against Adversarial Attacks: Evasion and Defense using BIM Algorithm

2022 6th International Conference on Intelligent Computing and Control Systems (ICICCS), Journal Year: 2023, Volume and Issue: unknown, P. 1381 - 1386

Published: May 17, 2023

Deep neural networks, such as ResNet50, have shown remarkable performance in image classification tasks. However, susceptibility to adversarial attacks, where small perturbations input images can result misclassifications, is a concern. The BIM algorithm popular technique for generating examples. objective of this research explore the vulnerability ResNet50 DNNs trained on ImageNet Stubs dataset evasion attacks using algorithm. Additionally, effectiveness various defense strategies, including training, defensive distillation, and transformations, examined. results reveal that are vulnerable BIM-based these defenses enhance their robustness. Overall, work underscores importance defending against ensure security reliability DNNs.

Language: Английский

---

Cyberattacks Detection Through Behavior Analysis of Internet Traffic

Procedia Computer Science, Journal Year: 2023, Volume and Issue: 224, P. 52 - 59

Published: Jan. 1, 2023

Network intrusion detection systems (NIDS) are actually used to detect suspicious activities such as viruses, shellcode, XSS, CSRF, worms, etc. There two types of the NIDS: signature-based and anomaly-based. Recently, Deep Learning have emerged promising techniques for classifying network attacks. In this paper, we propose a method analyze traffic behavior through classification using features. The results indicate that Multi-Layer Perceptron (MLP) Convolutional Neural (CNN) achieved similar performance with 94% accuracy when all features in dataset. However, use feature selection XGBoost, Pearson correlation, mutual information, models slightly lower 91%, but these demonstrate effectiveness methods enhancing by reducing complexity removing irrelevant

Language: Английский

---

Mitigation of HTTP Flood DDoS Attack in Application Layer Using Machine Learning and Isolation Forest

International Journal of Electrical and Electronics Engineering, Journal Year: 2023, Volume and Issue: 10(10), P. 6 - 19

Published: Oct. 31, 2023

Distributed Denial of Service (DDoS) attacks, specifically HTTP flood DDoS have become a constant and substantial threat to online companies critical services due the growing popularity web-based applications technology. attacks inundate web servers with an overwhelming volume seemingly legitimate requests emanating from compromised devices or botnets. Traditional mitigation approaches, often reliant on rate limiting traffic filtering, struggle discern between malicious traffic, leading service degradation downtime. Methods for identifying abnormal behaviour involve gathering preprocessing data, generating features, developing Isolation Forest algorithms. The power this method comes its ability detect anomalies in real-time, making it easy identify block attack traffic. As such, is significant feature methodology. In tandem Forest, machine learning empowers system adapt proactively emerging vectors, enhancing resilience face evolving threats. This research presents novel approach fortify application layer against by utilizing techniques, central focus algorithm. experimental validation results show that proposed framework can effectively recognize mitigate minimal interruption false positives. tests were run benchmark datasets KDD Cup 1999 NSL-KDD, stated here enhance basis model enable achieve objective.

Language: Английский

---

SmFD: Machine Learning Controlled Smart Factory Management Through IoT DDoS Device Identification

Published: Jan. 4, 2024

To prevent a website, network, or device from operating, Distributed Denial of Service (DDoS) attacks transmits large amount data to it. This attack makes use "botnet," which is an enormous collection pilfered devices that simultaneously transmit massive requests and the target system. In smart factory management, where lot are linked each other via Internet Things (IoT), DoS could be very risky. IoT essential factories, but these hacks have ability make them useless, might unfavorable effects. Downtime serious problem because it prevents (IoT) working, slows down production raises costs. DDoS may employed as diversion riskier behaviors compromise security, such unauthorized access breaches. Additionally, corruption loss occur, harming business's reputation long-term operations. proposed model ML trained chip systems capable real-time analysis. They identify patterns typical activity immediately anomalies indicate attacks. These not only trigger alerts, they also assist in identifying compromised devices, enabling prompt efficient action safety measures. The can manage new threats continually adapting learning things. building's managers security personnel see on basic screen. this research study, four distinct methodologies were used. Each provided unique method for approaching challenges related machine categorization. XGBoost, K-Nearest Neighbors (KNN), Logistic Regression, Gaussian Naive Bayes among techniques investigation's conclusions XGBoost stood out top performer continuously produced best results showed exceptional performance throughout range tasks assessed.

Language: Английский

---