search

networked-systems-iith / AdaFlow

AdaFlow: An Efficient In-Network Cache for Intrusion Detection using Programmable Data Planes
MIT License
0 stars 1 forks source link

Cache Contention #9

Closed Sankalp-CS21MTECH12010 closed 1 year ago

Sankalp-CS21MTECH12010 commented 1 year ago

Rebuttal Comments We picked CIC-IDS2017 dataset as it had about 14 different types of malicious flow attacks. We agree that more traffic \textit{might} change the results – we hypothesize that more traffic would impact precision and bandwidth overhead, but not recall as the cache design prioritizes malicious flows – we can get results by camera ready.

To-Do:

  1. Communication bandwidth comparison with *Flow on CAIDA traces.
praveenabt commented 1 year ago

@Sankalp-CS21MTECH12010

the todo items seems incorrect. First, we need to know avg pkts/sec and avg flows/sec (traffic characteristics) for each dataset we are aware of. Next, study the impact on Adaflow metrics (precision, recall, bandwidth overheads) as avg. pkts/sec or flows/sec increases.

Sankalp-CS21MTECH12010 commented 1 year ago

@praveenabt Since we are already evaluating for different datasets, can this trend not be extrapolated from the same?