This means that even the systems at the computing centre are starting to believe that I have managed to complete my studies. With graduation papers and my thesis safely in my bookshelf, I bid farewell to the Helsinki University of Technology. My official graduation day is 19.12.2007, so I managed to graduate just before the CS department I was enrolled in disappeared in a puff, to be replaced by the "Faculty of Information and Natural Sciences". I'm glad that I managed to get out when I did, before everything else broke. :)

The seven and a half years I spent at HUT were punctuated by great feelings of success and even greater feelings of frustration. It has changed me as a person in many ways, for the better and also somewhat for the worse. A lot of understanding was gained, at the expense of some idealism and youthful naïveté. All part of growing up, I suppose.

Work on the thesis went OK. I got the first complete version out in June, then spent a few months first waiting for feedback and then taking the feedback into account, and finally managed to get it ready, approved, printed, cased in and graded. I'm pleased with the result I got, even though there still was a little room for improvement. But, with perfection being the worst enemy of ready (or adequate), some tradeoffs had to be made.

Given the amount of effort spent in my studies I can heartily agree with what a friend once said: "HUT is a hard place to get into but it's even harder to get out of there".

Now it's time for me to consider my options for my future. This will take some pondering, but stay tuned!

A lot has happened again. Last week was a bit hectic, with me and my boss writing a paper about the things we've done so far. The conference is in Canada in the summer, here's hoping that the paper is accepted.. :) I also started to write my thesis, instead of just hacking on code. Not much text yet, but I'm trying to create a more-or-less sensible TOC first. After that is done, it's just filling in the blanks, right?

Next up: ordering a bunch of cards for me to destroy. Or, hopefully not destroy, I'm trying to add an attenuation pad to the cards so that I get the transmission power down to an acceptable level.

P.S. My new MacBook Pro is here. 15" widescreen, 2.33GHz Core 2 Duo, 2 GB of RAM and other miscellaneous goodness. Yay. :)

P.P.S. Muistinnollaus (the anniversary of Tietokilta, my study guild) was fun. :)

I am stupid (or something else is stupid, I'm not exactly sure). Anyhow, I managed to prove today that researchers need to get out of their chambers into the real world more and all their problems will be solved.. My problems with the transmit power attenuation were not caused by the cards, drivers or myself not functioning properly. Not at all.

They were, however, caused by one not too well documented feature of the cards I'm using. They have, like basically any WLAN card has, an AGC (Automatic Gain Control, for those of you not properly indoctrinated in acronyms) circuit on the receiving end right after the antenna. This is used to limit the received signal's dynamic range to a more sensible level from the 70dB required (-80dB to -10dB) by the WLAN specification. This is all well and good, as it allows a much cheaper A/D converter to be used on the card, leading to cheaper cards for all of us and other nice pink fluffy things. The AGC is probably a simple analog one, designed to work during the short preamble in the beginning of every WLAN transmission. It does not need to be too accurate or fine-tuned, it just needs to compress the signal's dynamic range to something requiring as few bits as possible to digitize.

What is not that well and good is that the gain/attenuation level of the AGC is not indicated anywhere. Thus, what happens is that if the signal received by the card is "strong enough" (for the AGC to be operating (either by attenuating it down or not at maximum amplification anymore, I cannot know)), the received signal level (RSSI) reported by the card, measured after the AGC, has no relation to reality whatsoever. This in turn means that the signal strength of WLAN station A measured by station B a few meters away has absolutely nothing to do with the transmission strength of station A. Goody. And here I am, banging my head on the wall/keyboard/monitor, wondering why the signal level jumps around funkily having my two test computers comfortably right beside each other. And it only took me nearly a week to realize what was happening.

I leave it for you to decide who was stupid...

I knew the trouble had to come someday..

I finally got around to measuring the actual power output of the cards. What I found out was that the power level fluctuates quite much on its own. The problem is that the transmit power level setting doesn't seem to work. It may well be that my kernel modifications don't work as they should (even though I doubt it), but what should be working is the Wireless Extensions txpower option. Alas, it doesn't seem to be working either. I don't know what the problem is but I'm planning to find out.

On a (much) lighter note, I passed the course on Discrete Mathematics! The grade was appallingly bad, but it's still passed and I'm not complaining about that :) It could be said that what I knew, I knew well, as the points per assignment were (6 0 0 6 0 6)...

Okay, I'm officially slow at updating this...

I was on a winter vacation last week, downhill skiing at the slopes of Hochjoch and Silvretta Nova in the Montafon Valley in Schruns, Austria. Not a bad way to spend a week, I must say. There was a good snowfall on the day we arrived there. The weather for the rest of the week was nothing short of unbelievable, with temperatures from about -3°C to +1°C and clear, blue skies for the whole week (except for one morning when the sky was partly cloudy)... Given the weather and location, pseudo-realtime-posting of images from the slopes (I love the Internet and modern technology in general :) was a good way to ruin cheer up the day of friends working back home... Evidently it is also possible to IRC while waiting for the chair lift to take you back to the top of the slope again.. :)

On the work side, things are progressing about as I think they should be. I'm now writing the userspace app to handle the actual key negotiation / verification. I haven't run into any problems yet, but they are sure to come. My supervisor asked me today about when I would be ready to run some sort of a demo with the software. This came as a surprise, but I'll try to manage at least something in a few weeks' time.

So, in a nutshell, work in progress.

Okay, it works! A bunch of bugfixes (some of which earned a "I am stupid" comment in CVS) had to be made, but after those stupidities were cleared out, the kernelspace modifications I did seem to work OK. But the hardest part so far is done. Yay!

Next up: Cryptographical key exchange (Diffie-Hellman probably), anti-man in the middle -things (need to check if there's any way to get the driver to report the signal strength / quality of received packets), and probably some privacy things with power modulation. Also need to find a way to measure WLAN output per packet to verify that the cards actually behave as advertised.

The proof-of-concept Socket API changes I made seem to be OK. I added a flag which allows me to set the radio interface into a low-power mode. The nice thing about the WLAN drivers I'm using is that they (seem to) allow setting the xmit power per packet. I'm trying to do a kludge which would allow an application to request a low-power mode per socket, to keep the range of transmitted packets to a minimum.

What I need to do next is to make sure that the flag gets moved down to the driver level properly. For that I'll need to install the modified kernel into some laptop or desktop which has a suitable WLAN card installed, and see that a) the driver really gets the parameter and b) that it honors the power setting. Satisfying a) should be easy, b) might need some heavy tweaking. Well, I can always walk to the EE department and see if they'd let me borrow a spectrum analyzer... ;)

On a completely different note, my Powerbook G4 12", a trusty companion for the last four years, is in the process of falling apart. First of all, it lost its battery completely now, powering off immediately when the power cord is disconnected. A more worrisome issue is that the machine seems to kernel panic when set on a tabletop, if one is not extremely careful. Evidently something is loose somewhere, but debugging that doesn't sound too tempting.. Therefore I'm now faced with a hard decision - MacBook or MacBook Pro. A 12" MBP would be an optimal solution, but Apple doesn't seem to want to produce one for me. Shame.

...for which I apologise. I've been getting busy with the implementation side of my thesis work, as my supervisor wants the results as soon as possible. I've spent the last week poking around the Linux kernel and MadWIFI driver code, trying to find the exact spots I need to modify. Yesterday I set up a QEMU-based test environment for myself, so I don't have to reboot my testing machines every time I tweak the kernel a bit.

My evenings have been more or less devoted to discrete mathematics. I have an exam on Saturday, and I've been trying to study for it for the past couple of weeks. It's not going good. Im feeling, well, if not depressed then at least quite downcast about the whole thing. I wish I'd done this course earlier, but that's what you get for being sloppy...

That's all for today, back to mathematics for me.. :/

Today I started working on my master's thesis at TML. As I suspected, my time was spent almost completely on practical matters (keys, access rights, user accounts etc.) I managed to do a very small bit of "real" work by ordering the WLAN cards I'll use on the work.

On my "normal" job, I've gotten used to starting projects and handling practical arrangements. Today's experience did not differ too much from that. There was the nice feeling of knowing everybody in advance, as practically all of the people I worked with for three years are still here. The feeling was very much like coming home. Worrisome, perhaps, but still kind of fun. :)

I'm back in Espoo again from my parents' house, where I spent most of my Christmas holiday. Eating, playing cards, lots of sauna and generally being lazy is a good way to spend Christmas.

A lot has happened in the year that's just ending. Last spring and summer, I worked on some really interesting projects (including a 1.5-week visit to San Francisco on March). I also went to Dublin and Prague with a couple of friends. Another high point of the year was the Assembly 2006 event in Hartwall Arena, at which I worked as a volunteer (on the AV technology crew). The event as a whole and also on my part went remarkably smoothly. As I worked practically full-time for the whole spring and summer, not much happened with my studies.

In the autumn I took study leave from work and tried to concentrate on my remaining studies, with pretty good success. Tying up loose ends is really good on the soul. Some very nice things also happened in my personal life, but I'll leave details on that for some other time.. :)

My outlook for the coming year is quite positive. The master's thesis project seems interesting, with graduation looming at the end of that tunnel. After my studies are done, I can take a fresh look on my life and what I would like to do with it. A lot of avenues are open or can be opened with some effort. This is also a very nice feeling.

A blessed and happy new year for all of you!

That's all folks!
...at least for this autumn term. Last exam for the term was today. It could have gone better, but I'm not complaining as my original idea was just to go and see what the exam is like. Now it's entirely possible that I'll pass.

The exam was divided into two parts, theoretical (essay questions) and "practical" (calculations etc.). At least 5 points are required from both to pass the exam. The theoretical part went more than fine, but the mathematical part might be a bit iffy. I think (and hope) that I'll pass anyways.

Now I should be gathering my stuff ready so I could leave tomorrow, but I think I'll play around with Flight Simulator for a while before that.. :)

I just read the paper on which my thesis will be based on. It seemed quite interesting, even though it was (purposefully) vague on technical implementation details - those will be my task.

Evidently my brain is operating within specifications, as I got at least some notions on what I have to take care of. First of all, there were a few questions which arose while reading throught the paper (one related to an attack scenario involving an attacker with a higher-gain antenna, the other to the actual handshaking protocol to be used). This somehow led me through principles of radio communication to thinking about the nitty-gritty implementation details. It seems that my brain jumps around likes to do free association at this time of day..

Implementation details I need to decide early on:

• Kernelspace - userspace separation.
  Do I dump more or less the entire logic into the kernel (as a loadable module, perhaps)? Or should I only add interfaces which I need (and which aren't there yet) and have the logic in userspace? Or something completely different?
• How to do the actual communication from kernelspace to userspace? Char device or /proc entry? Pros/cons for these?
• How to order the WLAN chipset around (I need to influence radio layer settings, which _should_ be covered by the Wireless Extensions, but you never know...)

Other issues currently on my mind: Exam tomorrow on automation and control systems. This means a large bunch of Fourier and Z transformations, transfer functions, Bode and Nyqvist graphs, fuzzy logic and whatnot. Wish me luck! :)

I received a phonecall a couple of hours ago, stating that the Telecommunications and Multimedia lab (will be referred to as TML from now on) at the Helsinki University of Technology (HUT) will hire me to do my master's thesis for half a year, starting on January 2007.

I will tell you a lot more about the scope of the thesis (and the larger project that it will be part of) as soon as I find out more. As far as I understand it now, it is related to creating and maintaining trust relationships between devices in an ad-hoc WLAN (or bluetooth) network. The researcher I spoke to had some neat ideas on how to accomplish this, and I also seemed to have some ideas on how to attack it, which I hope will lead me to some ideas on how to make the whole thing secure.

What this boils down to is that the study leave I've been on for the last three months will continue for at least the next six months. I hope that my colleagues at Nixu are not too disturbed..