Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore data poisoning attacks through interactive network visualization in this 40-minute conference talk from Ekoparty Security Conference. Discover how adversarial attacks on machine learning models can be visualized and understood through graph-based representations, moving beyond traditional detection methods to interactive exploration of poisoned training data. Learn about GraphLeak, an open-source web tool that transforms training data into visual networks, allowing real-time interaction with both clean and compromised datasets. Examine how data poisoning alters network structures through notable examples and tested case studies, while gaining insights into the visual patterns that emerge when training data is manipulated. Witness a live demonstration of a data poisoning attack, from execution through data collection to visualization using GraphLeak. Understand how network science concepts apply to cybersecurity, particularly in detecting tampered data through structural analysis. Gain practical knowledge about data provenance and poisoning phenomena in machine learning contexts, with applications for both offensive security researchers and cybersecurity professionals seeking to identify and mitigate adversarial data manipulation attacks.
