2021-shapira-flowpic
FlowPic: A Generic Representation for Encrypted Traffic Classification and Applications Identification
canonical link → · doi: 10.1109/TNSM.2021.3071441
Abstract
Represents unidirectional flows as two-dimensional histograms of packet size
and normalized arrival time, then applies a convolutional neural network to
distinguish traffic categories, applications, and encryption settings.
Team notes
Expanded version of 2019-shapira-flowpic. Sections III-B and IV describe
overlapping 60-second blocks; later experiments also consider shorter
blocks. The representation is not confined to connection handshakes.
Figures 1 and 2 illustrate application and traffic-category patterns.
This is dataset-based classification research, not evidence of a particular
censor's deployment, and it does not evaluate REALITY or XTLS Vision.