
FlowLens classifies flows directly on programmable network switches efficiently, targeting security applications. To cope with the limited hardware resources, we propose aggregation strategies calibrated with Bayesion optimisation. For covert channel detection FlowLens enables a 150 fold increase in monitoring capacity with an accuracy drop of only 3% vs collecting full packet distributions.
FlowLens is yet another uPVN use case, specifically targeting network security. FlowLens was joint work with our colleagues Diogo Barradas, Nuno Santos, and Luis Rodrigues at IST. It has been accepted for presentation at the next Network and Distributed System Security Symposium (NDSS). Stay tuned for Diogo’s presentation of this very cool work in Feb 2021!









