Course: CS125
Name: Daniel Meredith
Abstract Due Date: 02/16/00
Submit Date: 02/16/00
Journal Ref: RFC2402
In continuing with the theme of my abstracts this semester, I read another
Request for Comment from my IETF working group. I have been following the IPsec
working group this term and reading one of their RFC per week. In last week's
abstract I outlined and commented on ESP,the first of the the two main
protocols that will make IPsec secure. This week I will examine the second
protocol, Authentication Header, or simply AH.
Authentication Header is a method of providing connectionless integrity and
data origin authentication for IP datagrams. It also can implement protection
against replay packets, if selected in the Security Association negotiations
between the two host or security gateways. One item that is noted up front in
the RFC is that AH cannot protect all headers on a packet due to the fact that
they may be altered by switches and routers in transit. Thus the protection
provided to the IP header by AH is somewhat menial.
The header field of AH are very similar to those of ESP. Most notably the
inclusion of both the Security Parameters Index (SPI) and the Sequence Number
for both AH and ESP provide a method of redundant checking on both the validity
of the Security Association for the session and the ordering of the packets.
The AH header also includes a large space for the Integrity Check Value (ICV),
which is basically and very complicated check-sum. This field is one of the
many areas that must be padded to meet a 32-bit or 64-bit multiple.
Like ESP, AH must be run in either transport mode or in tunnel mode. Tunnel
mode is similar to the idea of IP tunneling, or running IP packets inside of
IP packets. This mode is more secure due to the ability of ESP to encrypt the
entire packet and then send it inside another packet. Unlike most of the IPsec
headers, AH must be placed in the proper order in a packet, and if it is used
in a Security Association all other IPsec headers must be placed in a certain
order, with AH being the first.
One of the largest limitations of AH is its inability to handle fragmented
packets. If a host or gateway receives a packet from a Security Association
that is using AH and the packet is fragmented, the packet must be dropped. This
could be a real issue on packets that move from networks with large maximum
packet sizes to those with smaller maximum packet sizes. The use of AH in this
type of transfer would make the entire transmission null and void. Therefore
the system must implement some method in the determination of the Security
Association to conclude whether or not the current transmission path will cause
fragmentation of the packets. If this is the case, then AH must not be used in
the Security Association. ESP is able to handle fragmentation quite well and
would therefore be a better choice in this situation.
Basically AH is a method of packing all of the Security Association, packet
order, host origin and authentication data all into one header so that if the
entire packet is then encrypted, the information for primary validation is
available without decrypting the whole packet.