List,
I am trying to apply some digital signal analysis methods to IPv4 networks.
Specifically, I have had some success using the Fast Fourier Transform in R to
find periodic events in IPv4 network traffic by producing periodograms. I store
network traffic in a mysql database so I have been using RMySQL to generate a
data frame of [timestamp,events] (the FFT is run on the vector of events).
Generating a periodogram in this case tells you that you have some periodic
event occurring on your network (such as a keepalive signal from an instant
messaging program) but it doesn't tell you anything more. It would be
useful to figure out which events in the network log being analyzed are
responsible for the "peak" in the periodogram--e.g. which IPs are
involved in some periodic activity.
So, I think what I want to do is take a subset of the vector produced by the
FFT and go from that back to the initial data frame of [timestamp,events], from
which I can go back to the source data and pull out the IPs and so forth. How
can I do this? My roommate (an aero engineer) has recommended that I look into
implementing a bandpass filter but I'm finding the material to be way over
my head.
I'm at the limits of my expertise (I am not an engineer :) so if anyone
just wants to point me at the topics I need to study, that would be
fine--although pointing out some specific examples would be great as well.
Thanks,
Pete
---------------------------------
[[alternative HTML version deleted]]
