NetGuard IDS

NetGuard IDS is an end-to-end data science pipeline applied to a real-world intrusion detection dataset of 488,000 network traffic records across 80 features, aimed at distinguishing benign traffic from infiltration attempts. The pipeline starts with a full data wrangling pass: removing NaN values, detecting and handling duplicates (28.56% of the raw dataset), and narrowing the feature set down to 16 key columns most relevant to classification.

Exploratory analysis includes a Pearson correlation heatmap that surfaced a 0.9951 correlation between Packet Length Mean and Average Packet Size, flagging a redundant feature before any modelling began and preventing wasted model capacity downstream. A Welch's t-test on Flow Duration returned a p-value of 0.4736, showing that no single feature could reliably separate the two traffic classes on its own and motivating a multi-feature machine learning approach rather than a naive threshold rule.

The project treats statistical rigor as a prerequisite to modelling: every feature-selection and preprocessing decision is backed by an explicit test rather than intuition.