greeblTubeSlice
biomedicine
cloud2
class="attachment-slide-thumb wp-post-image" alt="biomedicine" />

Author Archives: e.rozier

Visualizing road orientation distributions

March 30th, 2014 | Posted by e.rozier in Uncategorized - (0 Comments)

I ran across some really interesting visualization work by Seth Kadish of Portland, Oregon whose blog, Vizual Statistics, showcases his work. Seth produces some excellent visualizations of these distributions for many major cities in the US, and a few major cities around the world. This is a really great example of how intelligent visualization can … Read more

Our work with Fortinet has been featured in Miami Magazine’s new issue. The official website isn’t carrying the article yet, so I scanned in a copy to post in the meantime.

One of the issues we are investigating at the TRUST Lab, is the development of reconfigurable storage systems. We are currently trying to develop a better understanding of read workload characteristics in order to better build adaptable systems for read performance. We’re currently investigating the following traces which look quite promising: Wikibench SOS Project Traces … Read more

Big Data for Cities

May 31st, 2013 | Posted by e.rozier in Uncategorized - (0 Comments)

NPR has a good article today on Big Data for cities. http://www.npr.org/blogs/13.7/2013/05/30/187056297/what-big-data-means-for-big-cities While most of the article focuses on the positive innovations Big Data is likely to bring to our world (in another article they liken it to the Steam Engine of the 21st century), they spend very little time about the serious privacy concerns, … Read more

The Importance of Embedded Security

May 28th, 2013 | Posted by e.rozier in Uncategorized - (0 Comments)

A fantastic talk from the 29th Chaos Communications Congress from researchers at Columbia. The abstract from the talk: We discuss a set of 0-day kernel vulnerabilities in CNU (Cisco Native Unix), the operating system that powers all Cisco TNP IP phones. We demonstrate the reliable exploitation of all Cisco TNP phones via multiple vulnerabilities found … Read more

Anatomy of a Hack

May 28th, 2013 | Posted by e.rozier in Uncategorized - (0 Comments)

An interesting account on Ars Technica of the ability of hackers to crack seemingly difficult passwords: http://arstechnica.com/security/2013/05/how-crackers-make-minced-meat-out-of-your-passwords/ The culprit? The short answer boils down to two variables: the website’s unfortunate and irresponsible use of MD5 and the use of non-randomized passwords by the account holders. This is one of the big problems we’re trying to … Read more

NASA Formal Methods 2013 Live Stream

May 14th, 2013 | Posted by e.rozier in Uncategorized - (0 Comments)

The organizers of NFM have graciously made the live stream of the conference available for broader participation. Check it out here: http://www.ustream.tv/channel/nasa-arc

Sources of realistic virtual machines

April 4th, 2013 | Posted by e.rozier in Uncategorized - (0 Comments)

Part of the work we’re currently undertaking in the TRUST Lab is the study of cloud environments, privacy, and security. As always, one of the challenges we face is finding realistic sources of data. For the experiments we’re running we wanted to find sets of virtual machines representing virtualized applications running in a cloud setting. … Read more

Big Data: Sources for Real Test Data

February 26th, 2013 | Posted by e.rozier in Uncategorized - (1 Comments)

One of the major impediments to working in Data Science is finding good data on which to run experiments. Since we suspect, based on prior research we’ve conducted, that infrastructure for Big Data needs to be adaptive because the characteristics of the data to be stored can require very different strategies, how do can we … Read more

HMMs of User I/O Operations

February 24th, 2013 | Posted by e.rozier in Uncategorized - (0 Comments)

Below is an example classification of write and delete activities generated by our analysis. Two example HMMs are given below, for users 1, and 42, respectively. HMM state labels represent the average size of a write within a class. HMM transition probabilities are estimated from user data. Next up is working on classifying users by … Read more