August 06, 2008

_Electronic Signatures in Law_, Stephen Mason, 2007

Electronic signatures are now present in legal cases to the extent that while they remain novel, they are not without precedence. Just about every major legal code has formed a view in law on their use, and many industries have at least tried to incorporate them into vertical applications. It is then exceedingly necessary that there be an authoritative tome on the legal issues surrounding the topic.

Electronic Signatures in Law is such a book, and I'm now the proud owner of a copy of the recent 2007 second edition, autographed no less by the author, Stephen Mason. Consider this a review, although I'm unaccustomed to such. Like the book, this review is long: intro, stats, a description of the sections, my view of the old digsig dream, and finally 4 challenges I threw at the book to measure its paces. (Shorter reviews here.)

First the headlines: This is a book that is decidedly worth it if you are seriously in the narrow market indicated by the title. For those who are writing directives or legislation, architecting software of reliance, involved in the Certificate Authority business of some form, or likely to find themselves in a case or two, this could well be the essential book.

At 130 or so, I'd have to say that the Financial Cryptographer who is not working directly in the area will possibly find the book too much for mild Sunday afternoon reading, but if you have it, it does not dive so deeply and so legally that it is impenetrable to those of us without an LLB up our sleeves. For us on the technical side, there is welcome news: although the book does not cover all of the failings and bugs that exist in the use of PKI-style digital signatures, it covers the major issues. Perhaps more importantly, those bugs identified are more or less correctly handled, and the criticism is well-ground in legal thinking that itself rests on centuries of tradition.

Raw stats: Published by Tottel publishing. ISBN 978-1-84592-425-6. At over 700 pages, it includes a comprehensive indexes of statutory instruments, legislation and cases that runs to 55 pages, by my count, and a further 10 pages on United Kingdom orders. As well, there are 54 pages on standards, correspondents, resources, glossary, followed by a 22 page index.

Description. Mason starts out with serious treatments on issues such as "what is a signature?" and "what forms a good signature?" These two hefty chapters (119 pages) are beyond comprehensive but not beyond comprehension. Although I knew that the signature was a (mere) mark of intent, and it is the intent which is the key, I was not aware of how far this simple subject could go. Mason cites case law where "Mum" can prove a will, where one person signs for another, where a usage of any name is still a good signature, and, of course, where apparent signatures are rejected due to irregularities, and others accepted regardless of irregularities.

Next, there is a fairly comprehensive (156 pages) review of country and region legal foundations, covering the major anglo countries, the European Union, and Germany in depth, with a chapter on International comparisons covering approaches, presumptions, liabilities and other complexities and a handful of other countries. Then, Mason covers electronic signatures comprehensively and then seeks to compare them to Parties and risks, liability, non-contractual issues, and evidence (230 pages). Finally, he wraps up with a discussion of digital signatures (42 pages) and data protection (12 pages).

Let me briefly summarise the financial cryptography view of the history of Digital Signatures: The concept of the digital signature had been around since the mid-1970s, firstly in the form of the writings by the public key infrastructure crowd, and secondly, popularised to a small geeky audience in the form of PGP in the early 1990s. However, deployment suffered as nobody could quite figure out the application.

When the web hit in 1994, it created a wave that digital signatures were able to ride. To pour cold water on a grand fire-side story, RSA Laboratories manage to convince Netscape that (a) credit cards needed to be saved from the evil Mallory, (b) the RSA algorithm was critical to that need, and (c) certificates were the way to manage the keys required for RSA. Verisign was a business created by (friends of) RSA for that express purpose, and Netscape was happily impressed on the need to let other friends in. For a while everything was mom's apple pie, and we'll all be rich, as, alongside VeriSign and friends, business plans claiming that all citizens would need certificates for signing purposes were floated around Wall Street, and this would set Americans back $100 a pop.

Neither the fabulous b-plans nor the digital signing dream happened, but to the eternal surprise of the technologists, some legislatures put money down on the cryptographers' dream to render evidence and signing matters "simpler, please." The State of Utah led the way, but the politicians dream is now more clearly seen in the European Directive on Electronic Signatures, and especially in the Germanic attitude that digital signatures are as strong by policy, as they are weak in implementation terms. Today, digital signatures are relegated to either tight vertical applications (e.g., Ricardian contracts), cryptographic protocol work (TLS-style key exchanges), or being unworkable misfits lumbered with the cross of law and the shackles of PKI. These latter embarrassments only survive in those areas where (a) governments have rolled out smart cards for identity on a national basis, and/or (b) governments have used industrial policy to get some of that certificate love to their dependencies.

In contrast to the above dream of digital signatures, attention really should be directed to the mere electronic signature, because they are much more in use than the cryptographic public key form, and arguably much more useful. Mason does that well, by showing how different forms are all acceptable (Chapter 10, or summarised here): Click-wrap, typing a name, PINs, email addresses, scanned manuscript signatures, and biometric forms are all contrasted against actual cases.

The digital signature, and especially the legal projects of many nations get criticised heavily. According to the cases cited, the European project of qualified certificates, with all its CAs, smart cards, infrastructure, liabilities, laws, and costs ad infinitum ... are just not needed. A PC, a word processor program and a scan of a hand signature should be fine for your ultimate document. Or, a typewritten name, or the words "signed!" Nowhere does this come out more clearly than the Chapter on Germany, where results deviate from the rest of the world.

Due to the German Government's continuing love affair with the digital signature, and the backfired-attempt by the EU to regularise the concept in the Electronic Signature Directives, digital and electronic signatures are guaranteed to provide for much confusion in the future. Germany especially mandated its courts to pursue the dream, with the result that most of the German case results deal with rejecting electronic submissions to courts if not attached with a qualified signature (6 of 8 cases listed in Chapter 7). The end result would be simple if Europeans could be trusted to use fax or paper, but consider this final case:

(h) Decision of the BGH (Federal Supreme Court, 'Bundesgerichtshof') dated 10 October 2006,...: A scanned manuscript signature is not sufficient to be qualified as 'in writing' under §130 VI ZPO if such a signature is printed on a document which is then sent by facsimile transmission. Referring to a prior decision, the court pointed out that it would have been sufficient if the scanned signature was implemented into a computer fax, or if a document was manually signed before being sent by facsimile transmission to court.

How deliciously Kafkaesque! and how much of a waste of time is being imposed on the poor, untrustworthy German lawyer. Mason's book takes on the task of documenting this confusion, and pointing some of the way forward. It is extraordinarily refreshing to find that the first to chapters, and over 100 pages, are devoted to simply describing signatures in law. It has been a frequent complaint that without an understanding of what a signature is, it is rather unlikely that any mathematical invention such as digsigs would come even close to mimicing it. And it didn't, as is seen in the 118 pages romp through the act of signing:

What has been lost in the rush to enact legislation is the fact that the function of the signature is generally determined by the nature and content of the document to which it is affixed.

Which security people should have recognised as a red flag: we would generally not expect to use the same mechanism to protect things of wildly different values.

Finally, I found myself pondering these teasers:

Athenticate. I found myself wondering what the word "authenticate" really means, and from Mason's book, I was able to divine an answer: to make an act authentic. What then does "authentic" mean and what then is an "act"? Well, they are both defined as things in law: an "act" is something that has legal significance, and it is authentic if it is required by law and is done in the proper fashion. Which, I claim, is curiously different to whatever definition the technologists and security specialists use. OK, as a caveat, I am not the lawyer, so let's wait and see if I get the above right.

Burden of Liability. The second challenge was whether the burden of liability in signing has really shifted. As we may recall, one of the selling points of digital signatures was that once properly formed, they would enable a relying party to hold the signing party to account, something which was sometimes loosely but unreliably referred to as non-repudiation.

In legal terms, this would have shifted the burden of proof and liability from the recipient to the signer, and was thought by the technologists to be a useful thing for business. Hence, a selling point, especially to big companies and banks! Unfortunately the technologists didn't understand that burden and liability are topics of law, not technology, and for all sorts of reasons it was a bad idea. See that rant elsewhere. Still, undaunted, laws and contracts were written on the advice of technologists to shift the liability. As Mason puts it (M9.27 pp270):

For obvious reasons, the liability of the recipient is shaped by the warp and weft of political and commercial obstructionism. Often, a recipient has no precise rights or obligations, but attempts are made using obscure methods to impose quasi-contractual duties that are virtually impossible to comply with. Neither governments nor commercial certification authorities wish to make explicit what they seek to achieve implicitly: that is, to cause the recipient to become a verifying party, with all the responsibilities that such a role implies....

So how successful was the attempt to shift the liability / burder in law? Mason surveys this question in several ways: presumptions, duties, and liabilities directly. For a presumption that the sender was the named party in the signature, 6 countries said yes (Israel, Japan, Argentina, Dubai, Korea, Singapore) and one said no (Australia) (M9.18 pp265) Britain used statutory instruments to give a presumption to herself, the Crown only, that the citizen was the sender (M9.27 pp270). Others were silent, which I judge an effective absence of a presumption, and a majority for no presumption.

Another important selling point was whether the CA took on any especial presumption of correctness: the best efforts seen here were that CAs were generally protected from any liability unless shown to have acted improperly, which somewhat undermines the entire concept of a trusted third party.

How then are a signer and recipient to share the liability? Australia states quite clearly that the signing party is only considered to have signed, if she signed. That is, she can simply state that she did not sign, and the burden falls on the relying party to show she did. This is simply the restatement of the principle in the English common law; and in effect states that digital signatures may be used, but they are not any more effective than others. Then, the liability is exactly as before: it is up the to relying party to check beforehand, to the extent reasonable. Other countries say that reliance is reasonable, if the relying party checks. But this is practically a null statement, as not only is it already the case, it is the common-sense situation of caveat emptor deriving from Roman times.

Although murky, I would conclude that the liability and burden for reliance on a signature is not shifted in the electronic domain, or at least governments seem to have held back from legislating any shift. In general, it remains firmly with the recipient of the signature. The best it gets in shiftyville is the British Government's bounty, which awards its citizens the special privilege of paying for their Government's blind blundering; same as it ever was. What most governments have done is a lot of hand-waving, while permitting CAs to utilise contract arrangements to put the parties in the position of doing the necessary due diligence,. Again, same as it ever was, and decidedly no benefit or joy for the relying party is seen anywhere. This is no more than the normal private right to a contract or arrangement, and no new law nor regulation was needed for that.

Digital Signing, finally, for real! The final challenge remains a work-in-progress: to construct some way to use digital signatures in a signing protocol. That is, use them to sign documents, or, in other words, what they were sold for in the first place. You might be forgiven for wondering if the hot summer sun has reached my head, but we have to recall that most of the useful software out there does not take OpenPGP, rather it takes PKI and x.509 style certificate cryptographic keys and certificates. Some of these things offer to do things called signing, but there remains a challenge to make these features safe enough to be recommended to users. For example, my Thunderbird now puts a digital signature on my emails, but nobody, not it, not Mozilla, not CAcert, not anyone can tell me what my liability is.

To address this need, I consulted the first two chapters, which lay out what a signature is, and by implication what signing is. Signing is the act of showing intent to give legal effect to a document; signatures are a token of that intention, recorded in the act of signing. In order, then, to use digital certificates in signing, we need to show a user's intent. Unfortunately, certificates cannot do that, as is repeatedly described in the book: mostly because they are applied by the software agent in a way mysterious and impenetrable to the user.

Of course, the answer to my question is not clearly laid out, but the foundations are there: create a private contract and/or arrangement between the parties, indicate clearly the difference between a signed and unsigned document, and add the digital signature around the document for its cryptographic properties (primarily integrity protection and confirmation of source).

The two chapters lay out the story for how to indicate intention in the English common law: it is simple enough to add the name, and the intention to sign, manually. No pen and ink is needed, nor more mathematics than that of ASCII, as long as the intention is clear. Hence, it suffices for me to write something like signed, iang at the bottom of my document. As the English common law will accept the addition of merely ones name as a signature, and the PKI school has hope that digital signatures can be used as legal signatures, it follows that both are required to be safe and clear in all circumstances. For the champions of either school, the other method seems like a reduction to futility, as neither seems adequate nor meaningful, but the combination may ease the transition for those who can't appreciate the other language.

Finally, I should close with a final thought: how does the book effect my notions as described in the Ricardian Contract, still one of the very few strong and clear designs in digital signing? I am happy to say that not much has changed, and if anything Mason's book confirms that the Ricardo designs were solid. Although, if I was upgrading the design, I would add the above logic. That is, as the digital signature remains impenetrable to the court, it behoves to add the words seen below somewhere in the contract. Hence, no more than a field name-change, the tiniest tweak only, is indicated:

Signed By: Ivan

Posted by iang at August 6, 2008 10:44 AM | TrackBack
Post a comment

Remember personal info?

Hit preview to see your comment as it would be displayed.