Preventing Denial of Service Attacks on Reliable Multicast Networks

Abstract

Multicast is finding a lot of application in modern day networks and the Internet. There are various existing protocols that support the wide range of requirements demanded by these applications. If all the receivers in a multicast group are required to get all the packets at more or less the same time (i.e. synchronized reliable receiving), then the transmission rate of the source ends up being controlled by the rate of the slowest receiver in this group. Although, this is a requisite in some applications, it poses as a serious threat to the group. In other words, if one or more receivers were to artificially create a packet loss, then the source would be busy sending repairs and will consequentially slow down the overall transmission rate. This leads to a Denial of Service attack on the other group members.

The goal of this thesis is to suggest a mechanism to deter, if not prevent, the hostile receiver(s) from causing such an attack. We first study the problem with respect to a specific reliable multicast protocol, viz. Pragmatic Generic Multicast (PGM), by conducting experiments, which prove that PGM is also affected by the 'slowest receiver problem'. If the source can work out an optimum transmitting rate, we may be able to reduce the repair requests in the network and have a more stable system. To achieve this, we look at the possibilities and advantages of using an auction-based mechanism, such as the Generalized Vickrey Auction (GVA) to compute the optimum rate, based on the rate requests from the various participating receivers. We implement our mechanism in PGM and conduct experiments in order to compare its performance to that of the existing PGM protocol. Our results prove that for a network having malicious members, an appropriate auction-based mechanism complemented with policing stabilizes the source transmission rate and hence prevents a Denial of Service attack on other group members.