International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 03 | Mar 2025
p-ISSN: 2395-0072
www.irjet.net
Enhancing Cyber security Using an Enhanced Deep Learning Algorithm: An All-Inclusive Method for Threat Identification Mahesh S Markad1 1Head of Computer engineering Department Samajbhushan Eknathrao Dhakane College of Engineering Shevgaon
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - In cybersecurity, threat detection is examining
businesses. Given the intricacy and scope of today's cyberthreats, traditional cybersecurity. Methods such as signature-based detection and heuristic analysis have fallen behind. This challenge emphasizes the need for innovative strategies that can enhance the ability to identify threats and provide preemptive measures. DL, a subset of AI, has become a powerful tool in cybersecurity, able to accurately and successfully detect and eliminate cyberthreats.
network traffic and system behavior to find and address possible cyberthreats such malware, phishing scams, and unauthorized access. Organizations may proactively reduce risks, preserve sensitive data, and guarantee the robustness of their digital infrastructures through effective threat detection. New and changing threats, like as zero-day assaults, are difficult for traditional threat detection systems to identify. Their efficacy is diminished since they frequently generate a significant number of false positives and negatives. Furthermore, because these techniques depend on wellestablished patterns, they are sluggish to adjust to new cyberthreats. Using the ResNet101 model to improve threat detection with the STIN dataset, this work explores the possibilities of deep learning (DL) for cybersecurity. Improving threat detection accuracy, reducing false positives, and guaranteeing prompt reactions to new cybersecurity threats are the main objectives of this study. With an accuracy of 96.24%, precision of 94.28%, F1 score of 96.47, and recall of 98.67%, the ResNet101 model demonstrated outstanding performance.
By employing sophisticated neural network architectures, DL algorithms are able to analyze vast amounts of data, identify intricate patterns, and identify anomalies that may indicate illegal activities. Unlike earlier methods, DL models can continuously train and adapt to new attack patterns, which makes them incredibly effective in a threat landscape that is always evolving. offering a comprehensive strategy to cyber security in order to enhance the threat detection DL algorithms. DL models must be optimized by modifying hyper parameters, improving feature extraction techniques, and implementing novel architectures in order to get higher performance. By applying optimization approaches, the proposed approach aims to strike a compromise between detection accuracy, computational efficiency, and real-time response capabilities. The paper also looks at the challenges of implementing DL-based cybersecurity solutions, including adversarial attacks, data quality and availability, and the interpretability of model output. These problems need to be fixed if AI-driven security solutions are to be reliable and trustworthy. The study also examines the importance of using a range of data sources and threat intelligence feeds in order to create comprehensive and full threat detection models. The following is a summary of this study:
These outcomes demonstrate how well the model detects and reduces cybersecurity risks. Furthermore, the ResNet101 model shows its potential for real-time threat detection applications by processing the STIN dataset in about 55 seconds. This paper offers a thorough examination of model optimization tactics, performance assessment methodologies, and dataset pre-processing approaches. The results advance our understanding of AI-driven cybersecurity and provide insightful information that practitioners and researchers may use to fortify digital infrastructure against advanced cyberthreats.
A powerful threat detection framework that utilizes the ResNet 101 architecture and the pre-trained model.
Key Words: Cyber security, deep learning, ResNet 101, threat detection, and intrusion detection.
Examines contemporary DL techniques in cybersecurity in detail, highlighting both their benefits and drawbacks.
1.INTRODUCTION
Standard performance measures are used to evaluate the performance of the pre-trained model and the proposed ResNet 101. The remaining sections are arranged as follows:
Cybersecurity has become a critical topic in the digital age due to the rapid expansion of digital infrastructures and networked systems, which exposes businesses to an everincreasing number of sophisticated cyber threats.
Section 2 provides a review of the literature evaluating earlier studies. The recommended threat identification
Cyber attacks, which can include malware infections, phishing schemes, and advanced persistent threats (APTs), pose a major risk to individuals, governments, and
© 2025, IRJET
|
Impact Factor value: 8.315
Section 3 provides a description of methods. Results and a discussion of the model's exceptional accuracy and performance metrics are provided in Section 4, and a
|
ISO 9001:2008 Certified Journal
|
Page 658