Explore a conference talk from the Hack In The Box Security Conference that delves into tracking fake news using deep learning techniques. Learn about a novel approach to identify attackers who spread misinformation for malicious purposes, such as manipulating stock prices. Discover how the researchers developed a deep learning model that maps articles to a Euclidean space, allowing for the identification of content and stylistic similarities between texts. Gain insights into the challenges of tracking anonymous or pseudonymous authors and the potential of this method to uncover their real identities. Understand the experimental setup involving a dataset of 100,000 articles and the model's improved accuracy in identifying new authors outside the training dataset. Benefit from the expertise of researchers from Baidu X-Lab, including their backgrounds in AI security, software analysis, object detection, and face recognition.
Overview
Syllabus
Intro
Short Attack
Fake News
Networks
Tracking
Performance
Unified Master
Subnet
Net Loss Function
Fast Texter
CAA Ingram
Summary
Novel Strategy
Dynamical Strategy
Experiments
Planning
Comparison
Contribution
Questions
Taught by
Hack In The Box Security Conference