The Haversian Canal

Netflix released an anonymized sample of their user database last year to aid in a competition to develop a better recommendation system. Bruce Schneier, a well-known American cryptographer, leads today's Cryptogram newsletter with a piece on how this and other anonymized datasets have been de-anonymized. In the case of Netflix, the set was compared against user reviews on IMDB. But it's not just the IMDB reviewers who were identified. 99% of the dataset was identified!

Anonymized datasets are certainly useful to researchers, but the social fingerprint is hard to smudge. Before releasing data into a networked world, leaders need to compare the nature of the information to the potential consequences of compromising privacy.

This article, and a number of the articles that it links to, may be of interest to those interested in technology in education.

quick summary of things that might relate to higher education:

1) The modes of learning and testing on laptops are perhaps not the same as traditional methods (IMHO: mathematics, including physics and what little math is done in chemistry and biology, is a remarkably robust example: derivations are still mainly done on paper or chalkboard.
2) Maintenance is nontrivial
3) The degree to which the learner has, and has had, access to computers in the past effects the relative value the learner places on the time they spent with a school-provided laptop. And it is possible a student can over-value their machine time.
4) The investment is measured in tens of thousands, hundreds of thousands to millions of dollars
5) design matters (keyboard size, screen size, 2 power supplies -- for home and school)

Imagine a world where you don't have ultimate control of the things you believe are yours. I'm not talking like a mortgage, I'm talking like 1984. Microsoft, Intel, and their minions have been working on it for a decade. They call themselves the trusted computing group. Here's a hacker's perspective on trusted computing. The trusted computing idea has roots at least as far back as 1997. The central idea is to put an encrypted co-processor, a Trusted Platform Module, or TPM, in every device, often called a Fritz chip after Senator Fritz Hollings, who sponsored numerous bills in support of the effort. Apple started deploying TPM motherboards in 2005,. The Trusted Computing Group's next members meeting is 26 to 28 February, 2008. I think their consistency over time and the timeline itself suggest just how serious these folks are. From the frontpage of the TCG website today:

The TPM is the root of trust that is the basis of the work of the other TCG work groups.

The Trusted Computing Group also has a conference, from 27 to 30 January, the Technosium (there's sickly pun in there). From one of the abstracts

The Trusted Network Connect approach provides a method to ascertain end-point integrity for clients seeking connectivity to a network. Through trusted network connection protocols and trusted platform mechanisms, platforms can be authenticated before being given full network connectivity. This presentation will address the Trusted Network Connect architecture, similar efforts and the role of the widely deployed Trusted Platform Module in TNC.

For more background on the implications of a Tyrannical Plutocracy Module*, here is Ross Anderson's Trusted Computing FAQ. Richard Stallman recommends calling Trusted Computing Treacherous Computing.

*Verbatim copying and distribution of the moniker Tyrannical Plutocracy Module are permitted worldwide, without royalty, in any medium, provided this notice is preserved.

A two year old boy presents with a rash for one day. Three weeks ago he fell and hit his right cheek on a sofa, causing a superficial scrape 2 centimeters across (see slideshow). He did virtually the same thing again two weeks ago (1 week after the initial injury). Six days ago his father documented an additional area of irritation at the right corner of his mouth (slideshow). Three days ago his mother brought him to a general pediatrician because the scrape was not getting better. A topical steroid was prescribed. Two days ago the boy vomited "until there was nothing left" about two hours after dinner. The parents the vomitus was only partially digested food and wasn't red, green, particularly mucusy, or full of anything that looked like coffee grounds. The boy vomited again after drinking a sippy cup of milk that night before going to sleep. The following morning the boy awoke with dry heaves. A rash behind his ears was noted, consisting of tiny flesh-colored to pink bumps. The rash continued to spread over his body for the next two days; the father reported the child seemed tired on the day of presentation but reported the boy had asked for and kept down water, then rice, and then red beans and rice. The child vomited after milk was reintroduced later in the afternoon on the day of presentation, after which the parents started giving the child Gatorade. The parents report a low-grade subjective fever and cough for the previous two days but no chills, night sweats, or changes in mental status. There is no history of seizures.

The child denied itching and pain in any bones or joints and denied headache. The parents report their son is potty training and his last bowel movement was on the morning of presentation and it well formed without blood. The child attends daycare and his vaccinations are up to date. His six-year old sister had Fifth disease approximately two months ago and both children had molluscum contagiosum 18 months ago, which spontaneously resolved approximately 12 months ago.

Examination revealed a lot of little bumps behind his ears, with some bumps over the cheeks and two areas of the right cheek where the bumps were so close together you couldn't tell where one ended and the next started (see slideshow). The eye and ear exams were deferred. A flat, scaly, 2 centimeter red patch was noted over the right cheek. There were a few lymph nodes felt below the jaw bone, but no lumps behind his ears and no spots in his mouth or the back of his throat. The neck was supple. The child did not attempt to itch the rash. Similar scattered tiny bumps were seen over the back, chest, and arms. There was no tenderness on when pressing firmly of the bones of the arms and legs and thorax.

What do you think it is? What would you do next?

(slideshow)

Andrew Vickers reports on the lack of data sharing among medical researchers in today's New York Times in Cancer Data? Sorry, Can't Have It. At first I thought encouraging patients to add the GNU Free Documentation License to their consent forms when agreeing to participate in the trial, but that wouldn't liberate the entire dataset. Institutional Review Boards could mandate the freedom of information, but at the end of the day IRBs really only do what they do under the force of federal laws, and the situation would remain in the same patchwork state it is now. If even federally funded projects refuse to release their information, the existing statutes seem to be insufficient. Would a FOIA request have standing? If not, it seems a FOIA request procedure for medical trials data would be a smart piece of legislation.

Pharma would predictably lobby against such legislation on the thoroughly false basis that it would put American researchers at a competitive disadvantage and make US federal dollars too burdensome to pursue.

Perhaps the best regulatory body is the same international group that has had success in the past: the International Committee of Medical Journal Editors. ICMJE is the body that mandated registration of all trials, before enrollment begins, for a trial to be considered for publication, and publication is required in the drug and device approval processes.

I have approaching a decade of hard-earned receipt-level data in Microsoft Money and my wife and I both know how to use it and we really are getting a better grip on our finances. I would rather use an open source program, but I'm not willing to flush all my data down the drain. A week ago Scott Carpenter wrote the most exhaustive review I've seen of KMyMoney and migrating from MS Money, and his advice is to kiss my old data goodbye. Well, if that's the state of things, I guess I'll have to wait until GnuCash, KMyMoney, or some other application figures out how to preserve the hundreds of hours my wife and I have put into this dataset. From the looks of it, KMyMoney is the better option in terms of interface, but the data migration is still a long way from where I need it to be. TurboCash bills themselves as pretty sophisticated, but one complaint I have is that they apparently only build for Windows. One of the reasons I want FOSS is because I'll have a much better shot of running it on Windows, Linux, and OS X. We use all three at home. Best of luck to these teams, I sincerely hope they figure this out sooner than later.

From his site.

“I am in this race to tell the corporate lobbyists that their days of setting the agenda in Washington are over. I have done more than any other candidate in this race to take on lobbyists — and won. They have not funded my campaign, they will not get a job in my White House, and they will not drown out the voices of the American people when I am president.”

— Barack Obama, Speech in Des Moines, IA, November 10, 2007

And, by-the-by, the code on his site is nice and clean.