The following is either explicitly taken from Stephen Mason (2007), Electronic Signatures in Law, Tottel, 2nd edition; or implicitly builds on that book.
A definition of a signature is:
a token of the intent of a person to authenticate and give legal effect to a document.
This is primarily a restatement of the evidentiary functions, below, but with the addition of two key parts: that there is a token, and that its creation or use signals an intent.
Traditionally, signatures have been described more by their function than by their form, or manifest objectivity. That is because their importance lies more in their function than their form, any token that might give evidence to a signing function can be accepted as a signature. Hence, let's look at function first.
There are many functions of a signature. To summarise Mason (2007 pages 21-22):
These functions are complex and varied, and the above only gives a taste. The mental trick here is to consider the the signature as no more than a mere token that is slaved to the wider functions above. Hence the function of the signature is far more important than the form of the token, and this is generally established early on in any protocol. This function then goes on to inform what form of token might be appropriate in the case at hand, but the reverse might not be true. (Cryptographers might recognise a trap-door function: the signing will inform as to the form of the signature, but the form will not necessarily inform as to the function behind the signing.)
With that in mind, we can now describe what it is that might be the manifest form of the token that makes up the signature. Classically, a signature is considered to be a customary and individual form of a person's name, inscribed in running-writing by the person on a document. However, this is just a custom, not a rule. In different cases, at varying times, the following have all been accepted as a signature before the courts (Mason, 2007 2.9 - 2.32):
These examples form a fascinating array of possibilities, which reflects the courts' preference to look at the function and wider protocol, rather than any mere token.
From that historical position, it should be no difficult leap to consider the following as potentially valid forms of signatures from the electronic world (Mason 2007 Chapter 10):
(Mason 2007) describes many cases where an email is signed by means as simple as a typed name. For example, in a case in Northern Territory, Australia, an email laying out a separation agreement ended with the words 'Regards, Angus' (Mason 2007 10.6). The judge applied the appropriate electronic signature law (s9 ETA(NT)2000) and said:
I am satisfied that the printed signature on the defendant's emails identifies him and indicates his approval of the information communicated, that the method was as reliable as appropriate and that the plaintiff consented to to the method. I am satisfied that the agreement is 'signed' for the purposes of s 45(2).
Where things get more difficult is why headers to an email would help to signify that a document is signed. In the cases listed (Mason 2007 10.21, 10.24), the courts leaned heavily on a 'mosaic' of emails that authenticated that an exchange had taken place (such as an offer and acceptance in contract law) and that the parties were aware of the import. Thus, the courts accepted that emails could be accepted as signed apparently on the basis of (a) a header including a recognised and familiar email address, and (b) participation in a wider context that made the function and purpose clearly indicate a conclusion of intent.
The same logic would apply to faxes and to telexes. Which leads us to an important conclusion on the form of a signature, as opposed to signing; you may take the above to mean that a header is a signature , and indeed the bullet list above suggests precisely that. That is the wrong conclusion. Instead, the courts generally concluded that the emails were signed, resting partly on the identification function of the header, but also on the intent found in the words. They did not designate or declare the header itself to be a form of signature.
Hence, it is possible to sign without a signature. In such cases, it can be suggested that any form of the signature is absent, as there is no token in particular. Once again we must thrash this horse; it is the act and function of signing the document that is at issue, not the form of any signature. Indeed there may appear no tangible or identifiable form or token that can be designated as a signature.
Equally important then is to investigate what forms have been found not to be a signature. Tantalisingly, Mason (2007, Chapter 2) drifts from signing across to sealing , or the use of a traditional seal. These older customary marks come in several forms, being wax with impression of some pattern, a pre-printed patterned paper circle pasted on a document, or a physical impression made over a document with a crimping tool.
Curiously, the use of a seal is separate and distinct to signing; it seems to be based on customs and laws that certain documents such as Wills & Last Testaments should be sealed as well as signed. Consider the following illustrative quotes:
"that sealing is signing, I am not convinced; for sealing identifies nothing; it carries no character ... and most seals are affixed by the stationers, who prepare the paper."
Sir John Strange, (Mason 2007 page 58).
"It is true that one piece of wax may serve a number of people, if each of them impress it himself, or one for all, but the proper authority, or in the presence of all, .."
Lord Denman CJ, (Mason 2007 page 60).
"Now, whether the mark is made by a pen or by some other instrument cannot make a difference, neither can it in reason make a difference that a fac-simile of the whole name was impressed on the will instead of a mere mark or X."
Sir C. Creswell, (Mason 2007 page 73).
Hence, sealing is not signing, nor is signing, sealing. Then, the presence or absence of the (impression of a) seal then is not enough for the court to decide a document was or was not signed (which, again, is a different question as to whether it is sealed).
A court will look for other clues to help that determination. For example, in the USA, use of a Japanese seal, or "chop", has been accepted as signing a commercial contract (Mason 2007 2.36). I would speculate (Mason does not) that the court incorporated the customs of Japan into its analysis, where the chop is traditionally used in the signing function. In contrast, a seasonal greeting paper seal, containing the words "Merry Christmas. American Red Cross, 1912 Happy New Year." was accepted for a will, as the seal was also manually inscribed with the initials of the testator (Mason 2007 2.36). The writing of the testator's initials, by pen, was evidence that the testator was intending to sign the document.
Note that I stress the case law on use of seals for two purposes: sealing is in Mason as not always being accepted automatically as signing (although cases went both ways), and because of the similarity between seals and electronic signature devices.
We can now move to create a set of requirements that capture the above. A good mechanism for signing would include these features:
In summary, the fundamental need is to understand and interpret the act of signing before any discussion of the form of signature can take place. By considering the functions needed, we can also understand why forms are so varied. By considering how courts aim to identify intent, and do not stress form, we can also consider how systems might be built to meet that goal.
Casting an eye to the above requirements, it can be easily seen then that electronic signatures, and their narrow siblings, digital signatures, only succeed well at the first requirement: to identify the signatory. Especially, digital signatures fail to establish intent in any reliable sense, and their ability to identify an entire instrument is easily broken (c.f., Ricardian Contracts). Finally, as form follows function, and as purposes of signing vary tremendously, severe doubt is cast on any one form being a catch-all or universal method.
Thus, contemporary technological discussions that discuss mechanisms of signatures are built on a foundation of sand. Treat with care any such discussion.
There endeth this poor scribe's attempt to define signatures and signing. Why do I need this? Other than the general joy of wisdom, I wish to examine whether a digsig can form part of the function of signing. Obviously, some people have sold this as a done deal; painfully, the english common law will likely have no truck with their intentions, as I hope is outlined above. It seems that the presence of a digsig will likely be ignored by courts in many cases, simply because it is poor evidence of intent.
That controversy aside, what should a wayward supplier of digsigs do? What would a CPS need to state if it were to rule on the use of digsigs in an evidentiary fashion? Is a digsig a signature and can its presence provide any useful evidence of intent? Or, are they mere "authenticators", cryptographically-sound evidence of documents unchanged, with no intent in mind? If they are not signatures, what could be used as signatures? And, how would you describe a protocol that would allow all of these things to work together?
Two Caveats: I'll change this article if better wordings turn up. Defining signatures and signing is a work-in-progress.
Secondly, this article assumes the English common law approach, and does not cover the European or civil-law approach. That should be done as well.
Posted by iang at July 20, 2008 07:01 PM | TrackBackHi iang
Interesting article. My experience of this sort of thing. In the past I have agreed the terms of the preparation of a fine chemical with a customer via email. This was a long(ish) email exchange which discussed specification, time scales etc. Curiously, even though I put my name at the bottom of the email I never considered that I'd 'signed it'. However, I was confident that a contract had been formed by going back to the basics of contract law and finding the three elements of:- agreement (offer and acceptance), intention to form legal relations (presumed 'cos it was a business environment), and consideration (I was making something for which I wanted to be paid). I regarded the signature as unnecessary evidence of formation of contract.
When I wanted a 'signature' I thought that the 'only' way of doing it was to use pgp or similar. I know better, now.
Posted by: darren at July 20, 2008 04:49 PMThanks for doing this and other articles on the foundations of signatures and similar issues. I think that technologists in this area are often guilty of ignoring three key points:
1. No human interaction stands alone; any given interaction is part of a larger pattern of interactions and context surrounding those interactions.
2. Human interactions are by their nature ambiguous; the presence of ambiguities is a feature not a bug.
3. Human have perfectly good mechanisms for coping with ambiguity.
Posted by: Frank Hecker at July 21, 2008 11:40 AMWhy not combine the proof-of-intent and cautionary functions of the user interface with cryptographic signatures? For example, starting with the document
"I promise to X"
a popup asks the user "do you agree"? If so, the software appends to the document the text
"I agree to the above, signed, Name"
[cryptographic signature and certificate(s) if any]
Alternatively, when the user types in "I agree with the above, signed, Name" into the email or web form, the email program can append that text and [cryptographic signature and certificate{s) if any] to the document.
Now as your article suggests, it is the "signed, Name" or the clicking of the "I agree" button that is doing 90% of the work here. The cryptographic "signature" is there simply in case a dispute should arise over whether the signature is a forgery. It helps prove (but does not by itself prove) that the signature was not a forgery -- it doesn't serve any other function. It doesn't prove, for example, that the user didn't hit the "I agree" button accidently, or that a virus didn't sign the document, or that somebody else didn't cryptographically sign the document using a stolen password or key. It does not by itself prove intent. But of course signatures without cryptography don't prove these kinds of things either, so the cryptography used as above, as long as it doesn't add to user inconvenience, doesn't make things worse in any way. (In both cases, BTW, it is the user interface with the human-typed or -clicked signature, and not at all the cryptographic signature, that is providing the cautionary function to the signing person.)
To see whether cryptographic signatures can play a useful role in forming legal signatures, one must ask a pure security question of the kind that Ian likes to ask -- are in fact such forgeries a problem, or are they likely to become a big problem? Of course they are considered a security problem in cryptographic theory, but are they a problem in real life? How many alleged contracts have actually been rejected, by courts or by the counterparty itself, because the alleged signer successfully argued that the signature was a forgery? (I do know of one big area where this occurs, namely in the use of stolen identities to sign up for credit in another's name. But requiring cryptographic signatures when signing up for credit electronically might simply move the identify thieves back to filling out paper forms, rather than greatly reducing the problem of identify fraud).
BTW, even properly certified cryptographic signatures can prove only _some_ kinds of forgery and not others. They can't prove that someone stole the user's password and forged the signature. They can't prove that the user didn't hit the "I agree" button accidently. (I suggest requiring the user typing in their name instead of clicking buttons for this reason). They really just provide what I call "post-unforgeability", namely they only prove that the signature was not forged _after_ the message left the user's computer, assuming the key was not stolen. Again we need to look at real-life attacks to see which of these threats are the most important.
Very importantly, cryptographic signatures must be seamless to the user. If they make signing up for credit cards, for example, more difficult, this could easily outweigh any savings from reduced identify fraud, as credit applications will be reduced and the thieves could just go back to filling out paper applications.
Posted by: nick at July 25, 2008 04:48 PM