Swarm-based intelligent models for developing cybersecurity frameworks with IDS.

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Network Security

Background:

• The increasing sophistication of cyber threats necessitates advanced real-time monitoring systems.
• Existing intrusion detection systems (IDS) struggle with the volume and speed of modern network data.

Purpose of the Study:

• To design and implement a scalable, swarm-based intelligent model for real-time intrusion detection.
• To improve threat identification accuracy and minimize potential damages through timely anomaly detection.
• To develop a low-latency framework leveraging temporal data analytics for dynamic threat identification.

Main Methods:

• Utilizing a multi-layered framework for temporal pattern identification in time series data.
• Employing swarm-based Long Short-Term Memory (LSTM) for feature extraction and anomaly detection.
• Implementing an adaptive threshold mechanism to reduce false positives and a lightweight strategy for low-latency.

Main Results:

• The swarm-based LSTM model achieved 98.7% accuracy, 96.5% F1 Score, and 95.3% precision on the KDDcup99 dataset.
• Demonstrated superior performance and efficiency compared to vanilla LSTM, GRU, and Bi-LSTM models.
• The adaptive threshold mechanism reduced the false positive rate by 18%.

Conclusions:

• The proposed swarm-based intelligent intrusion detection system offers optimal scalability and efficiency for real-time cybersecurity.
• The system effectively identifies anomalies in high-speed data streams with improved accuracy and reduced latency.
• This approach provides a robust solution for dynamic threat identification across multiple domains.