End-to-end anonymisation of a synthetic healthcare dataset (55,500 records): GDPR attribute classification, HMAC-SHA256 pseudonymisation, a re-identification attack showing where pseudonymisation alone fails, then k-anonymity and l-diversity, closing with a measured privacy-utility tradeoff.

Plotly Dash app visualising global data breaches, with animated choropleth maps of victim countries and threat-agent origins. A two-stage pipeline (numbered clean/aggregate/merge scripts plus a BeautifulSoup scraper) backfills missing fields from the original breach reports.

Tableau dashboard of the Tradition–Secular axis from the World Values Survey, tracking how countries move along it from 1981 to 2022. Shipped as a reproducible pipeline: a notebook plus a data-free workbook that rebuilds the dashboard from your own WVS download, with the survey data never redistributed.

Hands-on study of where pseudonymisation breaks. Hashes identifiers with RIPEMD-160 and SHA-512, then breaks them back with brute-force and dictionary attacks; works through RSA and ElGamal cryptanalysis and the A5/1 stream cipher; and runs a chi-squared fairness audit on the results.

R workflow comparing k-means and DBSCAN (via OPTICS) on the Stat2Data Hawks dataset, evaluated with silhouette score, WSS, and WB-index. Custom staggered pair plots built with ggplot2 and patchwork; dependencies pinned with renv.