Stevan Harnad wrote:
sh> There are indeed powerful new ways of authenticating texts and dates in
sh> a digital archive. I hope my colleague, Adrian Pickering, will reply,
sh> describing his system, which CogPrints and Eprints may be
sh> implementing.
sh>
sh> But let me point out that at this time, priority and plagiarism are not
sh> the real problems, and I doubt they are what is delaying the inevitable
sh> era of self-archiving. (In context: getting people to read and cite our
sh> work is already such a problem that it is hard to imagine that there is
sh> much of a market for stealing it!)
Self archiving is, of course, something that engineers/scientists are
very familar with when it comes to keeping records for IPR purposes.
This model will continue as we progress towards entirely electronic
record keeping in this arena. Hence methods need to be adopted
which are amenable to this end.
There is only one emergent method, but with many variations. It is all
based upon hashing the digital document and then certifying (or
notarising) the hash. (The hash must be cryptographically secure.)
Digital time-stamp services are fast developing to cater for this.
However, all of these services have an 'Achilles Heel' - if the service
doing the time-stamping disappears or has its service compromised then
you lose the validity of your time-stamp (your priority date).
Assuming, however, that you can use varing and distributed methods to
fix your hash (and, thereby your document) in time, the authentication
then involves replaying the hash/time-stamp process and then showing
(in court perhaps) that this is identical to the one submitted at the
time. (As far as I am aware, no priority case has used this as a defense
yet, but it cannot be far off - particularly in the US, where electronic
transactions now have legal force.)
Accordingly, the original documents must be self-archived just s paper
ones always have been in the engineering/scientific arena (the
traditional log book). I already use hashing to link my own digital
documents to my log book. The only extra step of proof required is to
show that no document other than the one in question can have existed
at the time the log book entry was made. Log book entries already have
credance, so it is only the collision and cryptographic properties of
the hashing algorithm which are in question (and still to be tested,
legally). Those who know (RSA, NSA/FIPS etc) are pretty confident in
their algorithms despite serious attacks by the best decryptographers
in the world.
Where can Open Archives such as Eprints.org help? One of the properties
of any time-stamping service is that they are entirely indifferent to
the subject and the material. Indeed, the emerging IETF time-stamp
protocol - by definition - does not permit the agency to know the
identity of the requesters. Moreover, since it consists of
time-stamping hashes, it cannot know anything about the material
either. It is truly disinterested. However, it suffers from being
compromisable - that Achilles Heel.
Electronic archives - particularly those favoured by collaborating
communities, are not disinterested insofar as their own contents are
concerned. But they can do distributed cross-validation by (a) first
doing their own hashing and registering and (b) then securing that
information remotely by using one or more FURTHER, truly disinterested
service with which they have a dispassionate link. The more
distribution that goes on, the more resilient the scheme becomes. As
authors, we can accordingly feel more confident that priority dates can
be proved.
There are many e-time-stamp services about. However, there is quite an
interesting registry system operated by Surety and its affiliates (from
work originating at Bellcore). Visit
http://www.surety.com/. One still
cannot be 100% confident that they are 'Achilles Heel' proof.
But as part of a distributed network, they are just another
service that could be used for robustness.
At Southampton, we are developing a truly distributed and scaleable
approach, elements of which we may well incorporate in Open Archives
like Eprints.org and its friends. It is based on another principle -
that of the 'widely witnessed event' i.e. all the registrations are
publically viewable and thereby subject to scrutiny, which also helps
the case and increases resilience. As soon as we are ready to say more
about, it the 'white papers' will appear on
http://www.probity.org/
Adrian Pickering
Lecturer
Electronics and Computer Science
University of Southampton
Southampton
UK
+44 23 8059 2898
Received on Mon Jan 24 2000 - 19:17:43 GMT