International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 02 | Feb 2025
p-ISSN: 2395-0072
www.irjet.net
Advancing Cyber Security Through Optimized Deep Learning Algorithm: A Comprehensive Approach to Threat Detection Sumalatha M S1, Manju J2 1 Associate Professor, Department of computer science and engineering, Mahaguru Institute of Technology,
Kayamkulam
2Professor, Mahaguru Institute of Technology, Kayamkulam
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Threat detection in cybersecurity involves
techniques like heuristic analysis and signature-based detection have not been able to keep up. This difficulty highlights the necessity for cutting-edge approaches that can improve the capacity to detect threats and offer preventative defenses. A subset of AI, DL has emerged as a potent instrument in cybersecurity, with the ability to identify and eliminate cyberthreats with high precision and effectiveness. DL algorithms can evaluate enormous volumes of data, spot complex patterns, and spot abnormalities that can point to criminal activity by utilizing complex neural network structures. DL models are extremely effective in a constantly changing threat landscape because, in contrast to previous techniques, they can continuously learn and adapt to new attack patterns [1].
identifying and responding to potential cyber threat, such as malware, phishing attacks, and unauthorized access, by analysing network traffic and system behavior. Effective threat detection enables organizations to proactively mitigate risks, keep sensitive data, and ensure the resilience of their digital infrastructures. Traditional threat detection techniques struggle to detect new and evolving threats like zero-day attacks. They often produce high false positives and negatives, reducing their effectiveness. Additionally, these methods rely on known patterns, making them slow to adapt to emerging cyber threats. This study investigates the potential of deep learning (DL) for cybersecurity by employing the ResNet101 model to enhance threat detection using the STIN dataset. This study focuses on enhancing the accuracy of threat detection, minimizing false positives, and ensuring rapid responses to emerging cybersecurity risks. The ResNet101 model achieved exceptional performance, with an accuracy of 96.24%, precision of 94.28%, F1 score of 96.47, and recall of 98.67%. These results highlight the model's reliability in identifying and mitigating cybersecurity threats. Additionally, the ResNet101 model processes the STIN dataset in approximately 55 seconds, demonstrating its suitability for real-time threat detection applications. This study presents a comprehensive analysis of dataset pre-processing techniques, model optimization strategies, and performance evaluation methods. The findings contribute the knowledge in AI-driven cybersecurity, offering valuable insights for both practitioners and researchers employing to strengthen digital infrastructure against sophisticated cyber threats.
In order to improve the threat detection DL algorithms, presenting a holistic approach to cybersecurity. To attain better performance, DL models must be optimized by adjusting hyperparameters, enhancing feature extraction methods, and incorporating innovative architectures. The suggested method seeks to balance detection accuracy, computing efficiency, and real-time response capabilities by utilizing optimization techniques. The study also examines the difficulties in putting DL-based cybersecurity solutions into practice, such as data availability and quality, adversarial assaults, and the interpretability of model results. For AI-driven security solutions to be dependable and trustworthy, these issues must be resolved. In order to develop thorough and complete threat detection models, the study also explores the significance of utilizing a variety of data source and threat intelligence feeds. The summarization of this study are as follows:
Key Words: Threat Detection, cyber security, ResNet 101, deep learning, intrusion detection.
1.INTRODUCTION
Due to the fast growth of digital infrastructures and networked systems, which exposes enterprises to an everincreasing number of sophisticated cyber threats, cybersecurity has become a crucial area in the digital age. Businesses, governments, and individuals are at serious risk from cyberattacks, which can range from advanced persistent threats (APTs) to malware infections and phishing schemes. Because of the complexity and magnitude of contemporary cyber threats, traditional cybersecurity
© 2025, IRJET
|
Impact Factor value: 8.315
An effective threat detection framework that makes use of the pre-trained model and ResNet 101 architecture. Offers a thorough examination of current DL methods used in cybersecurity, pointing out their advantages and disadvantages. The suggested ResNet 101 and pre-trained model performance are assessed using a standard performance metrics.
The following is an arrangement of the remaining sections: A review of the literature assessing previous research is given in Section 2. The suggested threat detection
|
ISO 9001:2008 Certified Journal
|
Page 402