leakage-detection

Benchmarking Benchmark Leakage in Large Language Models

An open workbench for leakage-aware, calibrated, explainable risk modeling on longitudinal EHR-style data — built for students, labs, and health-informatics research.

A data quality auditing library for time-series tabular data with data-driven leakage detection.

This repository contains the coding aspect of the project Leakage Detection in Smart Water Distribution Systems, where the collected data is used for training an ANN model to detect leakage.

The AutoMLQuantILDetect package utilizes AutoML approaches to accurately detect and quantify system information leakage. We also provide different approaches to estimate mutual information (MI) within systems that release classification datasets to quantify system information leakage.

A guided data-science lifecycle for Claude Code: frame, explore, baseline, validate, model, evaluate, and report — with leakage and honesty checks built into every stage.

Repo contains codes implemented for different side channel attacks and faults for CS579 (physical attacks and countermeasures) for my Master's at Oregon State University

Agent-native, local-first AutoML for tabular data. Automatic leakage guardian, reproducible recipes & native MCP server for AI agents. 🐈⚡

Personal Information Tracker (PIT) for Android Applications, Detecting Information Leaks from GUI Components (e.g., EditText, RadioButton, Spinner)

Published in Artificial Intelligence in Health (2026) AIH. Leakage-resistant ML for cardiovascular risk prediction in diabetic patients. DOI: 10.36922/AIH025490111

CLI that applies the Clever Materials falsification test to materials-science ML datasets: detects confounding via a bibliographic fingerprint (year/author/journal leaking the target). Built on Jablonka 2026 (arXiv:2602.17730) + Huang&Cole / Nandy datasets (Scientific Data).

Research Artifact for ASIA CCS'25 Paper: *BranchGauge: Modeling and Quantifying Leakage in Randomization-Based Secure Branch Predictors*.

This project was presented at the AWS Impact Plus Hackathon held at Strathmore University. The project was aimed at Water Conservation to counter the effect of Climate change.

AI-powered leakage detection platform built with DBT workflows, anomaly detection, and predictive analytics.

End-to-end churn prediction project using the KKBox dataset

This is me learning how to quickly improve the quality of my models.

Open, low-cost monitor that warns factory workers when a machine frame becomes electrically live (STM32C031)

Early-warning model for the "Great Decoupling": search pages that keep their impressions but quietly lose clicks. Random Forest vs. a transparent baseline on 123,683 pages from a 79M-row warehouse, with client-grouped validation and a leakage audit.

Arduino Home Assistant Water Flowmeter (and Leakage WatchDog)

Timing side-channel analysis of leaky and constant-time STM32 quantum-control firmware using DWT cycle timing, Saleae PA0 pulse widths, TVLA/dudect-style Welch t-tests, effect size, mutual information, and attack success metrics.