The primary goal of our research is to detect rogue APs
from a central location (a switch that supports a subnet) with
the detection independent of the wireless technology. We
show a scalable solution, thus not attempting to reassemble
data before analysis. Also, this solution will function
independently of the signal range of the rogue APs. Our
research involves comparing traffic characteristics of flows
from different sources in a LAN segment and detecting traffic
coming from a wireless AP.
Our scheme starts with the hypothesis that a wireless link
in a network path of multiple links would cause a more
random and temporally different spreading of packets, as
compared to a path that has only wired links. Consider Figure
1 on the following page. The objective is to differentiate the
scenario shown in Figure 1, in which a switch port is
connected to a network segment that has no wireless links,
from the scenario shown in Figure 2, in which a switch port is
connected to a segment with at least one wireless link. The
assumption is that a majority of ports in a switch are
connected to network segments that have only wired links.
The processing and decision making are performed at the
switch with the input as the link layer traffic traversing, in
both directions, a switch port. The number of hops between
the switch and end point will most likely affect the temporal
characteristics of traffic as observed at the switch. Queuing
and congestion tend to mask the temporal shaping of traffic
through end points. However, we consider scenarios that
involve network segments with, at most, 2 links from the
detecting switch. Such scenarios are commonly observed in
most Ethernet local networks. The reliability of wired links
makes the temporal characteristics of traffic, in a path, to be