New Research on Performing a Side-Channel Video-Fingerprinting Attack

Four talented researchers from the DCU School of Computing worked on this piece, including Deborah Djon, Darragh Connaughton, Dr Geoff Hamilton, and Dr Andrew McCarren.

This research paper investigates the feasibility of a side-channel attack on encrypted YouTube video streams. Researchers used machine learning models, specifically Convolutional Neural Networks (CNNs), logistic regression, and Dynamic Time Warping (DTW), to identify videos based on patterns in HTTPS packet sizes. A CNN achieved the best performance, identifying videos with an 80% F1-score in an open-world scenario (where most videos are unknown). The study also contributed a large dataset of YouTube video network traces. The findings highlight a privacy vulnerability in encrypted streaming services and suggest the need for improved security measures.

Read the full paper here: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10733632