"Palladium" and TCPA

[Markus Kuhn]

Ross Anderson wrote on 2002-06-29 20:18 UTC:
> The mere existence of a mechanism that can be used for remote censorship is
> sufficiently bad that we should do what we can to stop it.

Is TCPA/Palladium/TrustNo1/whatever really a suitable mechanism for
remote cencorship?

If I was asked to implement a system for wide-area containment of
pirated content ("remote cencorship"), then I would not start with
targeting the end systems.

A cencorship system is far more effective at distribution nodes, because
network infrastructure is publicly visible (compliance with regulations
is far easier to verify than in end systems), and it is operated by a
smaller number of entities (a hand full of telco companies), which again
are far easier to control than the vast number of end users.

If Hollywood really understood what full-scale content protection looked
like, then they would lobby that the law-enforcement interfaces that
telecommunication service provider offer already today for eavesdropping
have to be extended to offer denial-of-service as well. This way,
Hollywood could get within hours court orders all over the planet to
shut down every website that starts hosting some new pirated content, it
can have access to specified files removed, etc. (Today's Internet
topology makes such court orders in the US de-facto probably sufficient
to control most of the planet.) Encryption won't help, because the
piracy is of concern only if a large number of potential customers can
access it, and if that is the case, then Hollywood can also access it
(encrypted or not does not matter) and identify, which parts of the
network infrastructure contribute to the public accessability of the
content, and then ask a court to shut down that network (as fine grain
as possible, how fine grain will depend on how much the network nodes
know about the application protocols they transport). Operators of
public file spaces will even have an incentive to provide fine-grain
denial-of-availability access to law-enforcement, because if that didn't
not exist, the block would be implemented at the next larger granularity
(e.g., block an entire set of servers or buildings as opposed to a
single URL or set of search keywords) and just be more annoying.

Law enforcement access to network traffic availability is a fairly
obvious extension of the already existing law enforcement access to
network traffic confidentiality. I'm actually surprised that we haven't
seen legislative proposals for that yet (apart from the emergency modes
of voice networks), but I am sure we will face such proposals as well in
the next few years.

> The EU copyright directive
> should mandate continuing access to protected content for legal deposit
> libraries, and we should see this reflected in the copyright regulations
> due to be published any time now.

That would indeed be nice. But why stop at just copy protection
mechanisms involving tamper-resistant hardware?

After all, compilation of high-level language source code is another
very widely used and relied upon form of content protection. Why not
amend legal copyright deposit requirements to demand deposit of the
entire documented source code and build environment of a product, as
opposed to the far less useful content of the shrink-wrapped box sold to
the end user. The deposit librarian must be able to verify that he can
recompile from scratch the deposited source code and ends up with
exactly the sold binary version before the author can enjoy the benefits
of full copyright protection. Or even make copyright for software
conditional to the deposit of a version that meets reasonably high
standards in terms of design documentation quanity (including high-level
structure, file formats, interfaces, etc.). If you really want to tackle
long-term preservation of software, then you have to preserve the
information necessary to make it feasible to reimplement interpreters
for digital content on future architectures, etc. If you can manage the
substantial additional workload for deposit librarians to check
compliance with such requirements, that is.

An open-source requirement for copyright deposit libraries sounds like a
very sensible and desireable feature of future copyright legislation to
me, especially when an acceptable way could be found such that pirates
can't abuse the system via the equivalent of inter-library-loan (which
is already today disabled for most deposit material). Copyright
libraries being allowed to give away only protected versions of the raw
content that they archive, until the copyright has expired, sounds like
a possible solution. All this is of course only feasible if there is
overall only a relatively a small number of build environments and
protection technologies.

> The temptation for software
> vendors to roll out revocation mechanisms, and the potential for these to
> be abused for censorship, are such that we will probably need legislative
> controls on document revocation.

Can you imagine a new type of abuse here, that is not already covered by
standard consumer protection laws?

I'd agree that classic consumer protection legislation has not been
written in mind with the ability of a product manufacturer to modify the
sold functionality after the product was handed over to the customer. If
I sold you a book printed in volatile ink that fades through oxidation
five days after you unpack it, you would cry out that this is fraud,
unless the seller very clearly warned you that this product is not a
normal book. The TiVo software update that removed the button to skip a
30 s advertisement slot conveniently was perhaps the most blatant issue
that suggests that consumer protection laws may have to be updated here
a bit. All that was already possible without any TCPA-like protection.
If you buy a functionality and expect (!) that it will last for a long
time, your vendor should not be allowed to take it away from you in the
future without your consent.

Again, I believe, not the wide availability of the mechanism (TCPA) is
the issue, but what consumer protection laws allow manufacturers to do
in principle, with or without any particular mechanism. The fact that
TiVo can remove the anti-advertisement button (remote cencorship) does
not mean that after-sale code updates (the technology they used to do
this) are in principle a bad idea. (They are incredibly useful for bug
fixing and proper extensions.)

> These are going to be messy. For example,
> the Fishman affidavit is forbidden in the USA, but published on websites in
> Holland and Germany. Will a US court be able to order its removal from PCs
> in Germany?

Well, a US court is de-facto already today able to suppress the access
to a German URL from Australia, by ordering the operator of the router
in the US to block it. They just haven't done it yet, because they
haven't understood yet what they could do if they wanted. No TCPA needed
for any of this.

For most consumer software purchases, there is anyway nationally sueable
entity representing the manufacturer. If I bought M$-Office in Germany,
I would buy it from Microsoft Deutschland GmbH under the full protection
of German law, and not from Microsoft Corp. under US law. Things get
only tricky for goods ordered directly in other countries.

The underlying problem is that nation states and their governments are
territorial entities, whereas telecommunication networks aren't. It will
take a long time until our political systems realize this and act
accordingly. The proper solution would be to treat cyberspace as a
domain with its own sovereign government and constitution. That's still
scifi at the moment, though ICANN etc. are very first (small and
difficult) steps towards that direction. Difficult, because the
interface and limits of control between a cyberspace government and a
territorial government still have to evolve. It can take a century or
more to sort that out.

With regard to TCPA, I think that it is in principle cool technology
that can -- with caution -- be used for many interesting things beyond
what it was originally designed for. For example, we discussed last
Friday the possibility to set up a mixmaster network, where the
integrity of each mix can be remotely verified by either communication
partner via the TCPA of the host on which the mixmaster runs, which
would make it substantially more difficult to take over mixmaster nets.
If TCPA were really as good as claimed, then even the owner of the
server should have great difficulties to get log data out of the
mixmaster that the software normally does not provide. Also, if the
manufacturers rolled out a PKI as part of TCPA, then you could, if it
became ubiquitous one day, start up a practical cryptographically
verifyable "Internet CPUocracy", that uses all sold TCPA machines as an
electoral register. Just to name two more crazy ideas which Bill Gates
might not yet have thought about ...

Markus

-- 
Markus G. Kuhn, Computer Laboratory, University of Cambridge, UK
Email: mkuhn at acm.org,  WWW: <http://www.cl.cam.ac.uk/~mgk25/>