TCP Congestion Control with a Misbehaving Router
Reviewer: Robert Dugas
The problem addressed is that of malicious hosts in a trust-based
congestion control framework.
This paper serves to identify and mitigate the significant risks posed by
malicious hosts (receivers) in current TCP deployments.
- Identifies various attacks to which current tcp in vulnerable
- ACK division
- DupACK spoofing
- Optomistic ACKing
- Proves significance of problem by demonstrating widespread
vulnerability as well as drastic consequences
- Proposes protocol alterations to mitigate/prevent attacks
This paper seems to represent the first recognition of trouble with the
trust based congestion control mechanisms utilized by TCP. Not only does
it underscore the significance of these such attacks, but also describes
the general shift toward the deployment of adversarial rather than
The primary methodology used by the paper is do enumerate various points
in the tcp congestion control protocol at which vulnerabilities can be
exploited. Testing is then conducted to show the drastic nature of the
flaws, and nmap probing reveals the extent to which vulnerabilities
The applicability of the problems and solutions in this paper are limited
only by users' inclinations to play by the rules. That is to say, there
are virtually no limitations.
The primary take-away from this paper I believe is to recognize the
sea change in thinking about transactions on the internet. No longer can
cooperation be safely considered the norm, instead one must assume malious
users and design accordingly.