New Technology - SecureBlue Wednesday, Apr 12 2006
friendly ai and technology 7:31 am
In the news this week is something extremely relevant to last week’s discussion on nanofactory regulation - on Monday IBM announced a computer security technology that encrypts data at the hardware level, something which has been done before but apparently not to this extent. The technology is called SecureBlue. A quote from news coverage on my.freeze.com:
There are multiple ways to achieve encryption, the mathematical art of encoding data to protect it from spying eyes. Specialized software can do the trick, as can hard-wired chips inside computers.
But IBM researchers contend that unless the encryption function is performed by a computer’s central processing unit, a supremely savvy hacker can tap into the pathway between the machine’s brain and the separate encryption engine.
To guard against that, IBM is announcing Monday that it has developed “SecureBlue” a set of encryption circuitry that can be integrated into any processor, regardless of its manufacturer.
“This thing is trying to be one of the most paranoid devices on the planet,” said Charles Palmer, IBM’s head security researcher.
One of the only times data is not encrypted within a SecureBlue chip is when it is displayed on the screen. We are rapidly approaching an era where the weakest link in computer security will almost always be the user (if we aren’t there already). IBM sees the technology being used in a variety of devices from PCs to handhelds and beyond.
The encryption scheme is not computation-heavy, barely consuming any overhead. IBM seems confident that the basic security of the technology will hold even if a hacker has ways of intimately monitoring the data streams within the hardware. Presumably access to IBM’s proprietary security algorithms is the only way to crack the code - and I would expect that the details of these algorithms are only known to at most a few hundred (more likely a few dozen) people at IBM’s research labs. Even with these algorithms, it’s not certain it would even be possible to decode any given package of encrypted data because it would be associated with a long, randomized string of bits (key). Perhaps the technology has safeguards so that data can be recovered even if the keys are lost? If so, a list of keys might be kept in a centralized location managed by IBM.
SecureBlue might be seen as a complement or successor to Trusted Computing.
It is fortuitous that IBM is the source of this new security technology, as it is also the company that built the computer that almost beat the highest-rated chess player on earth (Deep Blue), the fastest computer on earth (Blue Gene), and the largest attempt at a computer simulation of the mammalian brain (Blue Brain).
Hollywood and the big business behind proprietary software will be cheering for this technology, because it gives them another way to potentially prevent consumers from copying their movies, music, software, etc. I’m cheering it on for slightly different reasons, that is, the technology’s role in protecting us from future risks associated with totally unrestricted computers and software.
Sometime in this decade or the next, there will be a revolution in desktop manufacturing. This needn’t be in the form of nanofactories - it may debut as a relatively expensive machine that uses macroscale technology to shape plastic and electronics components into toys, tools, and simple gadgets. People will eventually be able to make custom products of high quality in their own home for low cost. The revolution is already starting to happen, with machines like MIT’s “fab lab” and the MCP Realizer, among dozens of others.
When copyrighted media such as songs or video clips get duplicated and distributed, the recording and film industries take a big hit. When expensive software like AutoCAD, Quickbooks, Windows XP, Photoshop, and Maya are copied, the software industry takes a big hit.
But these might be ultimately unavoidable. Information has a tendency to run free, and if security isn’t built into the foundations of the technology, it’s futile to stop the torrent by suing people one by one.
With desktop manufacturing, it will be a different story. If the design for a hot new product becomes public knowledge, then the value of the product will plummet mere days after its release, eliminating the motivation to both invent and invest. The dangers of a malfunctioning product will potentially be duplicated millions of times over. Without hardware-level restrictions and safeguards, performing a recall on a home-fabbed product will be neigh impossible. To make things worse, copycats will attempt to create similar products that circumvent safety restrictions.
This is just the beginning of why hardware-level encryption will be so much more important in the future than today. There are large classes of both existential risk and intense global nuisance that will be facilitated by insecure computing. These include bioweapon design, missile design, cyber-terrorism, remote control of military hardware, and much more.
The single greatest long-term risk of powerful insecure computing is probably self-improving Artificial General Intelligence (AGI) that is indifferent to human welfare. As available computing power increases, it gets easier to build an intelligent computer. (How much easier we don’t know.) However, it doesn’t get any easier to build an intelligent computer that cares about humans with the same complexity and subtlety that we care about each other - a must if you’re aiming for smarter-than-human AI. To oversimplify a bit, the former is a matter of trying things out until something works, and the latter is about developing a formal theory of what an agent will do given a starting set of preferences and the ability to reprogram itself recursively. Both will take a lot of brains, but creating any AI is a problem that lends itself to brute forcing much more than creating a certain type of AI.
Before we let powerful, unrestricted computers be available to just anyone, we should solve the problem of Friendly AI. A successful solution would give us allies who actually grew up in the world of code and will have a much better idea of which types of computation are truly dangerous and which are harmless - a question which humans are ultimately unqualified to answer.
April 12th, 2006 at 1:22 pm
I think what you wrote is wrong on many levels. First of all, “the details of these algorithms are only known to at most a few hundred (more likely a few dozen) people at IBM” is a classical example of security by obscurity. By itself there is nothing wrong with such, but only if this is not the only security measure. At least this is true with crypto algorithms. Moreover, an algorithm that is not published is prone to be not safe, just because of the lack of peer review.
Then, you touch the DRM issue. The problem why this hardware crypto won’t work for this purpose is because in the case of secure communication we are dealing with a sender, a recipient, and an adversary, while in the case of DRM the recipient and the adversary is one and the same entity; so it does not really matter how clever and bullet-proof an algorithm is - as soon as the content legitimately delivered to the recipient in some form, it can - and will - be stolen^H^H^H^H^H^Hcopied.
April 12th, 2006 at 5:18 pm
“When copyrighted media such as songs or video clips get duplicated and distributed, the recording and film industries take a big hit. When expensive software like AutoCAD, Quickbooks, Windows XP, Photoshop, and Maya are copied, the software industry takes a big hit.”
This is highly controversial. While they no doubt take *some* hit, a *big* hit is another story altogether. For example, Adobe has publicly claimed to lose $X to piracy, where X = (cost of Photoshop)(approximate number of copies downloaded), disregarding the question of whether or not those who pirated Photoshop would have otherwise bought it were piracy not an option.
In any case, whatever dangers of unrestricted computing are, preventing unrestricted computing will not necessarily have the net effect of improving our safety. The more I can do with my computer, the more valuable things I can do and the more wealthy both I and society become. Wealth increases the safety of society in innumerable ways.
Also, your claim that it is good that this is being done by IBM (presumably as opposed to, e.g., Microsoft or somebody else) strikes me as kind of silly. IBM may be able to do some brilliant things on the making-fast-computers front, but that does not qualify them to make good judgment about what level of freedom of use of computers is best for society (answer: total).
Finally, humans are absolutely qualified to determine what types of computation are harmless and harmful, or at least of the types that other humans are capable of creating. Building very secure computer systems is quite possible, or companies would not be able to make a living doing it. Friendly AI may be the security to end all security, but humans are quite capable of the security to end most security.
April 13th, 2006 at 1:41 pm
from TFA:
Bruce Schneier, founder of Counterpane Internet Security Inc., said more fully integrating encryption and processing would likely improve a machine’s performance. But he said it was “just stupid” to claim that hackers would otherwise target the transmission between a computer processor and a separate encryption engine.
Far more likely, he said, is for someone to try to steal data when it was unencrypted _ such as when it appeared in plain text on a computer screen.
“Security is a chain and it’s as strong as its weakest link,” he said. “They’re talking about taking a very strong link and making it a little bit stronger, at best. Maybe.”
/quote
Couldn’t have said it better myself.
So instead of getting at the data between the TPM and CPU, they’ll get at it between the CPU and the video chipset - if not on the screen itself.
It’s hype.
–Nato
April 13th, 2006 at 1:44 pm
From ToFA:
“The TPM is a step in the right direction,” Palmer said. “But it is not a bulk encryption device, and it would probably melt if you try to use it for an encrypted anywhere capability.”
So even IBM is selling this from a performance angle, and not from a security one.
–NAto
April 13th, 2006 at 1:58 pm
I hate to say it, but you have some dangerously naive views on patent and copyright, Michael. Hasn’t the Free Software movement provided strong enough evidence that IP incentives aren’t necessary to drive innovation?
http://n8o.r30.net/doku.php/whyipretardsinnovation
http://levine.sscnet.ucla.edu/general/intellectual/against.htm
April 13th, 2006 at 4:39 pm
Whether or not patents retard innovation is a contentious issue that is being argued worldwide. There is no decisive answer one way or the other, and there are extremely smart people on both sides of the argument.
The success of the Free Software movement does not necessarily imply that nanoproduct design will thrive in a patent-free economy. Programmers have spare time to design free software because their skills are in-demand and they get paid well when they aren’t writing code for no pay. Nanoproduct designers will not necessarily have an alternative if it turns out to be impossible for them to collect royalties from patented products. Perhaps a form of nano-socialism would ensue.
Also, remember I foresee nanofactories being a properietary platform. There will be no “Unix of nanofactories”. Those who manufacture the factories will ultimately have the final say about which kind of products can be built by their machines.
April 15th, 2006 at 5:09 pm
FOSS has come a long way from geeks working part time from basements in their underwear. Bruce Perens, among others, points out that most software written today is not “shrink wrap” written for copyright-enforced royalties. The shifting of applications to web-based service provider models is well underway, and will be well established by the time nanofactories arrive. The shrinking market share of shrink wrap (if you’ll pardon the pun) doesn’t seem to correlate well with developers salaries, nor with the avilability or development of FOS software.
http://perens.com/Articles/Economic.html
We do agree that the platform will likely be proprietary though. I don’t think it will be industrial intellectual property, though. I’m thinking it will be nationalized under the aegis of national security. Not that there’s a huge difference between industry and government these days, anyway. *shrug*
April 21st, 2006 at 12:59 pm
Michael & Nato: I, too, am a bit ambivalent on the roll of IP legal rules in the context of emerging tech. But, in terms of the socio-ecnomic and legal landscape, and in terms of persons having access to this tech (”a fleet of robots and a nanofactory or two for every household…”, in addition to “a chicken in every pot…”), we should take a serious look into Kelsonian and neo-Kelsonian ideas and legal institutions. In my judgment, they dovetail well with an otherwise more conventional classic-liberal socio-economic framework. Basically, it just involves tweaking existing finance law & institutions. See, e.g., Kelso & Adler’s classic *Capitalist Manifesto*, as well as the work of Prof. Robert Ashford of Syracuse U. Law School. See also *chapter 10* of James Albus & Alexander M. Meystel, *Engineering of Mind: An Introduction to the Science of Intelligent Systems* (Wiley-Interscience, 2001), wherein Albus discusses neo-Kelsonian proposals to widely diffuse the *ownership* (and thus real-income-derived-from) of robots & AI systems, as they come into the workforce. Now I don’t necessarily agree with every jot & tittle of the specifics of Albus’s proposals, but I think something broadly along these lines deserves at least to be carefully researched & considered, as we try valiantly to thread the needle between nanarchy (in the negative/problematic sense of that term) and nano-socialism, which is just as problematic (though in a slightly different way). After all, do we really want bureaucrats (whether federal or corporate…) doling-out robots and nanotech as they see fit? We’d likely wind-up with something horrible similar to Terry Gilliam’s **Brazil**!! Better to be as near genuine free-market-based (with proper [Kelsonian??] institutional set-up in place…) as possible, don’t you think?–if only for efficiency reasons… Also, something along Kelsonian/Albusian lines may help in terms of consumer-oriented (as distinguished from, say, military/police-oriented) R&D *incentives* (there’s that “i” word again…wink…)
Wonderful blog, Michael!! I’m so happy & honored to comment here…Keep up the good work…