International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 04 | Apr 2025
p-ISSN: 2395-0072
www.irjet.net
Threat Detection Using Dilated Convolutional Neural Networks 1Gunturu.Maheswarareddy,2Alaparthi.Yaswanth,3Janga.Vedeshreddy, 4Andagula.Trinadh, 5Mr. Y.V. Narayana, Assistant Professor. 1,2,3,4,5Dept. of Information Technology, VVIT, Andhra Pradesh, India
-------------------------------------------------------------------------***------------------------------------------------------------------------
Abstract
patterns by capturing long-range dependencies, thereby improving accuracy in threat detection. vulnerable.
With cyber threats like Distributed Denial of Service (DDoS) attacks becoming more sophisticated, protecting online services has never been more critical. Traditional threat detection systems often Depend on preset rules, which limits their effectiveness.at spotting new and evolving cyber threats, including Previously unknown attacks. To mitigate this, this paper presents a threat detection system powered by dilated convolutional neural networks (DCNNs), which are highly effective at analyzing network traffic patterns by capturing longrange dependencies. This allows for more accurate and efficient threat detection. The model is trained using supervised learning on a dataset containing both normal and unwanted traffic, with preprocessing and data normalization techniques applied to handle issues like data imbalance and noise. Performance is Assessed based on essential classification Indicators such as accuracy, precision, recall, F1-score, and ROC-AUC are used for evaluation, ensuring reliability in identifying Online threats. Testing results indicate tha DCNN-based system significantly outperforms traditional CNN and RNN models, achieving 99.03% accuracy, 98.88% precision, 99.03% recall, 98.95% F1-score, and 99.27% ROC-AUC. These results highlight the effectiveness and scalability of the proposed model in real-time threat detection, offering organizations a powerful tool to strengthen cybersecurity and proactively defend against emerging cyber attacks..
2. LITERATURE REVIEW In today’s digital world, cybersecurity is more crucial than ever, with organizations constantly battling increasingly sophisticated cyber threats, from the internet, especially Distributed Denial of Service (DDoS) attacks Traditional Threat Detection Systems (TDS) They mainly depend on established rules and recognizable attack patterns, which makes them less effective against emerging and evolving threats, such as unknown attacks To handle these challenges, machine learning (ML) and deep learning (DL) have been utilizedhave emerged as promising approaches in network security. Among these, Dilated Convolutional Neural Networks (DCNNs) stand out due to their ability to detect cyber threats by capturing long-range dependencies in network traffic. This section reviews existing research on ML/DL-based threat detection techniques and highlights the advantages of DCNNs over conventional models. 2.1 Traditional Threat Detection Approaches Threat Detection Systems (TDS) Are typically classified into two main categories: Pattern-Based TDS (PBTDS): These systems, like Snort and Suricata, rely on predefined attack signatures. Although they work well against known threats, they are unable to identify unknown attacks(Garcia et al., 2019).
Key Words:
Cybersecurity, Threat Detection, DDoS Mitigation, Network Security, Deep Learning, Dilated Convolutional Neural Networks, Anomaly Detection, Real-Time Security.
1. INTRODUCTION
Given these challenges, there’s a growing need for more intelligent and adaptive solutions.
With the rise in cyber threats, especially Distributed Denial of Service (DDoS) attacks, ensuring network security has become a major concern. Conventional Threat Detection Systems (TDS) largely depend on static rules, which are inadequate for identifying new and advanced attacks. To improve upon these limitations, this study introduces a novel TDS utilizing Dilated Convolutional Neural Networks (DCNNs) to effectively recognize both known and emerging threats. DCNNs are particularly effective in processing network traffic
© 2025, IRJET
|
Impact Factor value: 8.315
Anomaly-Based TDS (ABTDS): These systems use statistical analysis or machine learning to identify unusual traffic patterns (Santos et al., 2020). While more adaptive than SBTDS, they often struggle with high false-positive rates and evolving attack tactics.
2.2 Machine Learning-Based TDS Researchers have explored various ML techniques to enhance threat detection, including:
|
Support Vector Machines (SVM): Useful for classifying attacks, but resource-intensive when handling large datasets.(Wang et al., 2021).
ISO 9001:2008 Certified Journal
|
Page 15