The following text is copyright 1995 by
Network World, permission is hearby given for reproduction, as long as
attribution is given and this notice is included.
The value of sure and
obscure packets.
By: Scott Bradner
Two weeks ago Patrick
Bird wrote an article for this paper about the use of packet-level encryption
(*Packet encryption may bury security concerns* - nw Nov. 27 pp45) but did not
happen to mention the work of the Internet Engineering Task Force (IETF) in
this area.
In August the IETF
published five security related Proposed Standards (the first step in the IETF
standards process) as RFCs. These documents are products of the IP Security
Working Group in the IETF. They provide an overview of a security architecture
for the Internet Protocol (RFC-1825), and descriptions of a packet
authentication extension to IP (RFC-1826), a specific authentication mechanism
(RFC-1828), a packet encryption extension to IP (RFC-1827), and a specific
encryption mechanism (RFC-1829). Support for these security features is
mandatory in implementations of the next generation of IP (IPv6) which claims
to be standards-compliant. They are optional for the current version of IP
(IPv4) in use on the Internet and in private IP networks.
Authentication is used
to ensure that the receiver of information transferred across a network knows
that the origin of the information has not been forged and that the information
itself has not been modified during its travels. There are many applications in
which confidentiality is not required, yet it is quite important to be able to
verify that the data is legitimate. An example of this is the routing updates
within the Internet. The information being carried (the location of individual
networks attached to the Internet) is not secret but it could be quite
disruptive if someone were to deliberately insert false information.
At other times the
information itself should be kept hidden for example, when transferring credit
card numbers from a buyer to a merchant. The encryption extension is used in
these cases.
Some people would claim
that performing these types of security functions at the packet level is wrong
since application level security is far more efficient than packet level. It is
clearly better in terms of the amount of data transferred if an application can
support authentication and/or encryption on its own, but the presence of packet
level security permits the use of security-ignorant applications in a secure
way.
One area that requires
additional effort is key management. In order to use security of any kind both
ends of an electronic conversation must agree on the security methods and the
encryption keys to use for a particular interaction. Frequently the key
exchange is the weak link in the security system. The IETF IP Security Working
Group is currently working on key exchange systems and expects to propose one
quite soon.
The IETF security
functions can be done using separate external devices, as in Patrick's
description, by the network filewalls increasingly used by corporations when
connecting to the Internet, or by using software installed on the end systems
themselves.
A number of prototype
implementations are running now and IETF security features should be soon
available in commercial products.
The widespread
availability of the ability to support authenticated and confidential exchanges
over private data networks and over the Internet is critical to the future
growth of electronic commerce. Packet level security will be a key enabler in
this area.
RFCs can be retrieved by
using anonymous ftp from ds.internic.net in the directory rfc. The filename for
RFC-1825 would be "rfc1825.txt". A URL for the RFCs is
http://www.internic.net/ds/dspg1intdoc.html
Disclaimer: While many
of the people at Harvard are quite sure of themselves, they are not so sure
about others, so authentication could be useful, but these are my own opinions.