Wednesday, July 17, 2013

MS Faculty Summit Day Two

Day Two of the summit started with a keynote shared between Peter Lee and Jeannette Wing, who now seem to be sharing heading up Microsoft research.  (Rick Rashid is stepping down from his Chief Research Officer role;  there was a nice tribute to him, with Ed Lazowska providing a very nice homage to his development of Microsoft Research over the past couple of decades.)  I'd say their talk was a bit "rah-rah" for Microsoft, but it was also quite "rah-rah" for basic research generally and its role in developing computer science, so I wouldn't want to complain.  (With their positive, enlightened view on research and the compelling way that they can describe and present it, perhaps those two should instead be heading up some of the large government programs in charge of sponsoring research.  Oh, wait...)  In particular, I'd be remiss if I didn't point out that Jeannette specifically called out (with several slides) the recent theory work on "interlacing families" by Adam Marcus, Dan Spielman, and Nikhil Srivastava (see links here, or discussion here from Nikhil Srivastava).  Nikhil works for MSR India, so this was an example of MSR-univeristy basic research collaboration.  (Of course, Dan was Nikhil's advisor at Yale, so one might hope for a bit more exotic an example, but still, it's a nice example of basic research MSR supports.)    

The afternoon had a demo session -- a room full of demos from various MSR groups.  There was good one on networking that I liked (can't find a pointer, if someone sends I'll update), but most seemed focused on visualization and human-computer interfaces.  A few with Kinect, and a pretty interesting one that was based on touch-screen-type technology but was focused on your feet.  (Sensors would be embedded in the floor.  It could tell who was who by what shoes they wore;  it could detect motions like tapping, kicking, even just weight shifts.)  With a screen embedded in the floor you could play "virtual soccer".  A further prelude to our eventual holodecks. 

The "spam" session was the most entertaining.  Saikat Guha of MSR presented their work on tracking ad-click bots and related ad fraud.  Part of the work was focused on determining how much fraud there was and where.  Using that, they can do things about it.  They found a particular type of malware that shadows a user, and when the user does a search but doesn't click on an ad, the malware wakes up and clicks on an ad, at most once per day.  The behavior then looks like a real user, so it's hard to catch;  on the other hand, clicking an ad once per day is itself a noticeable behavior...

But the most entertaining talk of all was, unsurprisingly, Stefan Savage, who was talking about the "economics of cybercrime" -- how they figure out the "money chain" of the companies sending the spam e-mail selling drugs and illegal software, how they think about the weak points in the "money chain" (it's the banks), and how they've worked to give this information to law enforcement so that law enforcement is better equipped to take down spammers engaged in illegal activity.  Needless to say, Microsoft is interested and involved -- they don't like pirated copies of their software being sold. 

Thanks to Microsoft for inviting me.

Red-eyes are a killer.  (They weren't 25 years ago.  I wonder what happened.) 

1 comment:

Anonymous said...

The paper on 'Measuring and Fingerprinting Click-Span in Ad Networks' would be a nice practical paper for your end of the wire class. Thanks for pointing it out.