UNSW-NB15 and KDD99: Network Intrusion Detection Datasets for Feature Analysis

Nour Moustafa from UNSW Sydney examines the feature characteristics of the UNSW-NB15 and KDD99 datasets for Network Intrusion Detection Systems. The paper applies an Association Rule Mining algorithm for feature selection and evaluates classifiers on accuracy and False Alarm Rate. Experimental results indicate the replicated UNSW-NB15 features improve efficiency on the KDD99 dataset.

Use Cases

• Evaluating feature selection algorithms based on network packet features mentioned in the description
• Benchmarking classifier performance on accuracy and False Alarm Rate for intrusion detection
• Comparing the efficiency of feature sets from different network traffic datasets
• Studying the impact of irrelevant or redundant features on anomaly detection systems

Strengths

• Focuses on two established benchmark datasets (UNSW-NB15 and KDD99) for network intrusion detection
• Analysis includes specific performance metrics: accuracy and False Alarm Rate (FAR)
• Applies a defined methodology: Association Rule Mining for feature selection

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Last update date is unknown; freshness unverified

Provenance

Source

Nour Moustafa, UNSW Sydney

Collection Method

Likely involves analysis and feature replication between two existing network traffic datasets.