Tuesday, November 27, 2007

HMRC fiasco and risks

so i've heard two unbelievably ignorant comments by senior politicians in the wake of the fiasco where HMRC accidentally "lost" 2 CDs with 25M people's children names, addresses and parents bank account details

1. biometrics wont have this problem
ok so what if the passport office accidentally released the database with 60M people's biometric data? well, according to experts, it is crypted doing a 1 way function so you can't reverse engineer someone's Iris or fingerprint from the data.
But what if there is 1 single bug in the way they'v done this? just 1. d it can.
everyone's biometric is compromised, once and for all and for ever. game over. doh.

2. encryption would have meant the problem with the 2 CDs wasnt a problem.

ok so how was the NAO (or KPMG) going to read the data? magic? no, they had to have the leys (and password) too. SO how were those sent? securely? do we know they aren't the same key and password as are used for loads of other data bases inside HMG's various termianly ICT challenged departments? no we dont, nor do they.

anyhow, both these are irrelevant - the fact that junior (and therefore large numebrs) of staff have access to the entire database means that it is effectively open to all and sundry (as with polis databases) who can afford to find any bribably or blackmailable or just careless person in a large population of junioer clerks.

"all the eggs in one basket" appears to be a phrase that wasn't part of Darling's education (or browns).

Wednesday, November 21, 2007

HMRC incompetence beggars belief.

Everyone's aware that the HMRC (british tax and customs government agency) accidentally shipped a plaintext file on disks in the ordinary post with 25M people's national insurance, bank account and other personal information, in response to a request from the Govermnet audit office for some sample data.
Ross Anderson and Ian Brown (and others) appeared on Newsnight last night to very good effect, and it was quite clear that the government spokesperson failed to understand the real nature of the problem which is not the "one off" nature of the error in sending a file unencrypted (this is an operational error of fairly huge proportions) but is the fact that this is symptomatic of a government that allows low level staff the ability to even create a copy of the entire database - this should not be possible, by design. Any decent system of mission critical data will have methods to control the damage that can be done - have they never heard of the idea of "need to know"??? As was pointed out, the fact that they are so ignorant of the simplest principles of data based access control means that they are unfit to propose other systems in this space (ID cards, NHS spine, etc etc) since they will make the same, criminally negligent, mistakes.

I'd like to point out that if I now claim someone has withdrawn money from my ank account, or masqueraded as me using my NI (national insurance == social security ID), they have no course but to believe that it is from this leak. we could run a massive denial of service attack on the bank accounts of many people now by simply observing that ID is virtually worthless.

pathetic, no - more - someone should go to jail.

oh, another note: someone seems to think (brown?) that they've "lost" the records!!!
they havnt - dont they understand that they have still got them - what they have done is PUBLISHED the records, by making a copy widely available. when wil lpeople understand that "sending" a datum is not "sending" - it is making a copy, and transmitting the copy...and if you send to persons unknown, you are publishing.

Friday, November 16, 2007

anti-social networks

great article on how to kill off community networks - should be mandatory reading for comms #101 startup wannabees and software engineers...:) this has a lot of good other points about standards, APIs and how not to make innovative technology succeed.

on the other hand...

now everyone (i.e. cory doctorow) is on the case about the built in contradictions of social nets - apparently, they will evaporate - so this is a bit like a "hawking radiation" model of black holes- unfortunately the likely model is far more like diffusion (social diffusion and memes are well understood at least statistically) where basically people will slow down in atttrachtion towards each social net hub until there's a stable equilibirum established....

Wednesday, November 14, 2007

on the tip of your tongue....and recall yet takes so long

so you know when you are trying to remember something (name of film, actor in film etc) and you know you know it, and your conscious effort tells you you know it, but not what it is - indeed, sometimes, it seems, the harder you try, the less you are able to recall the item - so how does this work? then you get distracted (perhaps for a long time, but sometimes just for a shortish while) and then suddenly, in an inappropriate moment (sleep awakeinging in extremis) the datum/factoid comes flooding back


so that is weird right? I mean the neuroscience of the brain says its a set of synapses that are slow, but not that slow, organised into some sort of (very high degree) holographic, or swarm storage....so retrieval might be slow - each "hop" in the net needs lots of time by silicon standards, and you need quite a few hops...

I am guessing that the memory is also partitioned, and there's only so much resource between the clusters of memory items and those resources are used to statmux requests for retrieval - if some region is currently isolated because you are
busy using the "links" to it for soemthing else, then is just like "call blocking" in a telephone net - it takes a while...and then, when you have re-organized things due to task switching, the pathways are freed up and a queued request (call) gets thru - this also helps explain the patchy way memory loss works - if you actually lose "pinout" from a region, it might all be still holding info, but you simply can't route to it....without re-training some other component - a lot of this reminds me of large scale P2P systems where request routing is slow and complex and re-org of nodes (known as "churn" in systems like pastry or kademlia) leads to some possible loss of info, and certainly to potential long periods (compared with single node performance) of no access....

curt cobain said something about memories...and I dont have a gun, which wasnt true

Tuesday, November 13, 2007

so where do metro drivers live? start or end of line

and how does it affect the schedule where they live eh? thats what I want to know

Hands off is also an odd expression when you think about it....as is hands-on (experience) - surely brains connected would be more useful?

Friday, November 09, 2007

copy protection and payment

so hollywood screenwriters are on strike because they don't get extra payments when
productions are released on other media according to the bbc story on WGA . Irony, isn't it that when users copy DVDs and MP3s hollywood runs around getting the RIAA on the case to sue them, but when hollywood copies material written by authors and makes more money, it isn't prepared to share it iwith the original creators. Classic hypocrisy in big business in action. why should we have any sympathy at all? send them all to Iran (no, wait, Friends is the most popular soap in Iran - dont send them there!!!)

Thursday, November 08, 2007

animated evidence

there was this incident in london where a brazillian was shot dead (deliberately ) by
armed police shortly after the london suicide bombings....the chattering classes beat
themselves up about it a lot out of guilt, but the police have been largely let off
although recently found guilty collectively under health and safety law (bizarrely)

the bbc has assembled a fairly detailed flash animation of the events - see
http://news.bbc.co.uk/2/hi/in_depth/629/629/7073125.stm

it would be interesting to know if anyone uses tools like this in court cases -
if done right, it could simply jury's work quite a bit

Blog Archive

About Me

My photo
misery me, there is a floccipaucinihilipilification (*) of chronsynclastic infundibuli in these parts and I must therefore refer you to frank zappa instead, and go home