Data-Driven Cyber Security in Perspective-Intelligent Traffic Analysis.

Area of Science:

• Cybersecurity
• Network Traffic Analysis
• Machine Learning

Background:

• Traditional rule-based cyber attack detection is being replaced by machine learning approaches.
• Large datasets are crucial for training high-performance machine learning models in cybersecurity.
• Analyzing social and internet traffic is vital for identifying and mitigating cyber threats.

Purpose of the Study:

• To review recent analytical research on cyber traffic within social networks and the internet.
• To introduce and demonstrate a novel data-driven cyber security (DDCS) methodology.
• To explore the application of DDCS in classifying network entities and user activities.

Main Methods:

• Utilizing concepts of similarity, correlation, and collective indication for traffic analysis.
• Employing a data-driven cyber security (DDCS) framework comprising data processing, feature engineering, and modeling.
• Analyzing diverse network and social flows with characteristics like fixed-size messages.

Main Results:

• Machine learning models show outstanding performance in cyber traffic analysis due to large datasets.
• The DDCS methodology provides a structured approach to data-driven cybersecurity.
• Effective classification of network hosts, applications, users, and social media activities is achievable.

Conclusions:

• Data-driven cyber security is essential for modern cyber attack detection and defense.
• The proposed DDCS methodology offers a robust framework for analyzing complex network and social traffic.
• Further research is needed to address challenges and explore future directions in this evolving field.