Data Poisoning

Overview

Human Data Poisoning (Current):

Humans inject malicious data into training sets Static corruption - poison data doesn't adapt Requires human understanding of target model vulnerabilities Limited by human knowledge of ML systems

AI-Driven Data Poisoning (Emerging):

AI systems could generate sophisticated poisoned data for future AI training Dynamic corruption - poison data could be designed to activate under specific conditions AI systems have deep understanding of neural network vulnerabilities Could operate at massive scale across multiple training pipelines

<!-- ## Prerequisites Check the roadmap for recommended prerequisites for this topic. --> <!-- ## Resources Resources and detailed content will be added soon. In the meantime, you can: - Explore related topics in the curriculum - Join the discussion in the community - Contribute your own knowledge and resources --- *Note: This is a placeholder page. Comprehensive content is under development.* -->

Research Questions

Recommended Reading