Data Science & Machine Learning Pipeline Security
Adversarial attacks on neural networks, model stealing, training data extraction, and PyTorch security hardening.
Course Overview & Objectives
Foundational security for data science. Learn how machine learning pipelines are attacked in practice: adversarial evasion samples, model inversion attacks that leak private training data, and malicious pickle payload deserialization in model weights.
What You Will Master
- Generate FGSM (Fast Gradient Sign Method) adversarial perturbations against image classifiers
- Identify arbitrary code execution vulnerabilities in serialized pickle / PyTorch checkpoint files
- Audit machine learning training data for leakage of personally identifiable information (PII)
- Harden ML pipelines using SafeTensors and differential privacy libraries
Prerequisites
- Introductory Python and basic mathematics/linear algebra
Platforms & Tools Covered
Detailed Curriculum Modules
1 modules structured from foundational theory through complex adversarial execution.
Adversarial Machine Learning Fundamentals
White-box vs black-box threat models and gradient-based perturbation techniques.
Hands-on Virtual Sandbox Labs
Zero local hardware dependencies. Provisioned in cloud containers via browser terminal.
Malicious Model Checkpoint RCE Demonstration
Demonstrate how unvalidated pickle loads allow remote shell execution and convert to SafeTensors.
Faculty & Lead Instructor
Direct weekly instruction, live office hours, and code-review feedback.
Piya Kohli
Thread Security EducationAI Security Specialist
Specialist in adversarial robustness testing and secure lifecycle management for deep learning neural networks.
Frequently Asked Questions
Everything you need to know about scheduling, cohort admissions, and lab access.
Is this suitable for data science beginners?
Yes, step-by-step Jupyter notebooks guide you through both the math and the practical code.
Ready to Master Data Science & Machine Learning Pipeline Security?
Join the upcoming cohort. Seats are limited to maintain a high faculty-to-student ratio and rigorous sandbox feedback.