-
00 : 01
© Distribution of this video is restricted by its owner
Transcript
×
Auto highlight
Font-size
| 00:01 | Hello. I'm a slice and I'm to tell you a little bit about |
|
|
|
| 00:07 | research interests that I have. Uh am interested in many things. So |
|
|
|
| 00:17 | can consider me as somewhat omnivorous but and large one can distinguish security, |
|
|
|
| 00:27 | performance computing. So what I'm interested security is sand boxing, which allows |
|
|
|
| 00:36 | to protect environments. The applications here creating cloud computing viruses and worms. |
|
|
|
| 00:47 | then there is group authorization. Which interesting because authorization of individuals tends to |
|
|
|
| 00:56 | slightly different. It is not entirely with group group authorization, especially when |
|
|
|
| 01:04 | are a member of several groups. then their digital watermarking aspects. This |
|
|
|
| 01:13 | related to integrity of digital objects, security. You have impact based digital |
|
|
|
| 01:24 | for video. It can be also to audio. These are essentially techniques |
|
|
|
| 01:31 | verification and actual medication. A recent recently, I've been interested in pyramid |
|
|
|
| 01:40 | biometrics. Uh basically people have stated uh biometrics are an advance over passwords |
|
|
|
| 01:56 | to some extent they are and to extent they definitely aren't. What is |
|
|
|
| 02:03 | an insidious aspect of that is that biometrics once they have been lost can |
|
|
|
| 02:13 | be recovered. That's of course in distinct contrast to a password. If |
|
|
|
| 02:20 | lose a password, once you realize been lost, you just change the |
|
|
|
| 02:25 | and that's the end of that You can't do that with biometrics because |
|
|
|
| 02:31 | very directly tied to a person and aspects of an individual fingerprints, iris |
|
|
|
| 02:42 | and so on and so forth. you've lost those, they can be |
|
|
|
| 02:48 | . And so what we're trying to is to see whether we can paramilitarism |
|
|
|
| 02:54 | so we can get the advantage of combined with that of passwords. Uh |
|
|
|
| 03:04 | are other aspects that statistical data based . Statistical databases have become more important |
|
|
|
| 03:15 | the last 2 3 decades simply because have more and more data and more |
|
|
|
| 03:23 | more data with privacy concerns. And we have questions of statistical database security |
|
|
|
| 03:35 | really are still to some extent This is related to inference control. |
|
|
|
| 03:45 | can we prevent people from posing innocuous queries and infer from those innocuous statistical |
|
|
|
| 03:58 | , seemingly innocuous statistical queries, information is supposed to be confidential private. |
|
|
|
| 04:06 | are techniques that have been shown to quite well for averages of the values |
|
|
|
| 04:15 | there are things like medians, maximum group together usually as selective functions that |
|
|
|
| 04:24 | require additional work. I have other in security that relate to access control |
|
|
|
| 04:32 | systems, e voting and privacy and that is related even remotely to any |
|
|
|
| 04:41 | these topics. If you're interested in . Talk to me high performance |
|
|
|
| 04:49 | I have been involved for quite a in high performance computing. Uh My |
|
|
|
| 04:56 | at some point develop a good deal high performance seismic data processing software for |
|
|
|
| 05:05 | oil and gas industry. As a , I became far more interested in |
|
|
|
| 05:10 | output management because what we are dealing are very, very large data sets |
|
|
|
| 05:16 | that's not going to change. Even uh data storage devices become more |
|
|
|
| 05:26 | we still have problems with getting enough storage for all the data that we're |
|
|
|
| 05:35 | . So data are a fundamental problem it's not going to go away. |
|
|
|
| 05:44 | so as a result, input output is important. Uh I have argued |
|
|
|
| 05:53 | a long time that the current uh of relegating input out with management to |
|
|
|
| 06:03 | operating system is a misguided effort. problem is that the operating system essentially |
|
|
|
| 06:12 | very little about the program. A , compiler on the other hand, |
|
|
|
| 06:17 | a lot more about the program than operating system will ever know. And |
|
|
|
| 06:23 | we can apply certain compiler techniques to computations. So input output management is |
|
|
|
| 06:35 | so that in particular allows us to use techniques from vector ization and parallelization |
|
|
|
| 06:41 | particular based on dependence analysis. So gets basically down to memory management. |
|
|
|
| 06:52 | are other interests, I started life a formal language theorist. I'm still |
|
|
|
| 06:59 | doing slight work on that. I am interested in certain aspects of the |
|
|
|
| 07:08 | of hanoi, I have made some generalizations of the Towers of hanoi Two |
|
|
|
| 07:15 | to parallel moves have characterized with together students the worst graphs as far as |
|
|
|
| 07:24 | requiring most the maximum number of moves more practically, I am interested in |
|
|
|
| 07:34 | from algorithms to software. Lots of go wrong when you have a good |
|
|
|
| 07:40 | and then you translate that into um, other interests of mine or |
|
|
|
| 07:49 | aspects of computing, including privacy, crime and surveillance, both legal |
|
|
|
A simple tooltip for the link