Differences

This shows you the differences between two versions of the page.

--- ongoing [2011/01/17 17:25] – bil
+++ ongoing [2011/10/14 12:57] (current) – bil
@@ Line 1: / Line 1: @@
 ====== Ongoing projects ======
-====== Leadership Election Among Numbered Agents with Constrained Interactions ======
+====== Web Crawlers Behaving Like Humans: Are We There Yet!?  ======
-**Student**:  Stephen Voland
+**Student**:  Dmitry Drinfeld
-**Supervisor**:  Patrick Dymond and Michael Jenkin
+**Supervisor**: Natalija Vlajic
-__Description__
+**Required Background**: CSE4480 prerequisites
-This project is an examination of the efficiency with which numbered distributed agents are able to elect a leader, given certain constraints.  It builds on the paper 'Efficient Leader Election Among Numbered Agents', by R. Chaturvedi, P. Dymond, and M. Jenkin.
-When collections of agents are deployed to solve a problem it can be beneficial to establish a leader among the group. For certain applications it may be desirable to establish this leader in some static manner. For other situations it is desirable to have the group of agents establish the leader themselves. Chatuervedi et al. (2011) developed an algorithm that enables a group of identical agents differentiated only by their unique identification number to elect their own leader. Chaturvedi et al.'s work contained a very simple simulator for a specific algorithm. Ongoing work requires the development of a more sophisticated simulator of this collection of agents, in particular the development of a simulator that allows for systematic evaluation of certain properties of the collection of agents.  It is also desirable to perform a rigorous mathematical analysis of some of the more simpler variants of the existing work.
-The project will involve the creation of a simulator which will take parameters for the number of agents, their internal initial state such as arrangement in space, and a function to determine the probability of an interaction between two given agents based on their positions.  In the most straightforward case, all paired interactions would be equally likely, and thus the agents would have no associated positions.  Another case would have the agents spread randomly on a bounded plane, where the probability of a pair interacting would be inversely proportional to the distance between them.  A third would have the agents distributed among rooms, and would allow an agent to communicate only with agents in the same or an adjacent room.  The program will simulate a leadership election using the supplied parameters, and report properties of the simulation including how many interactions were necessary to complete the election, if the election was successful, and actual number of interactions taken for election
-Theoretical results for such elections will also be determined, and will be compared with the experimental results for mutual validation.
-====== UCOSP: Development for Encyclopedia of Life ======
-**Student**:  Feng Sun
-**Supervisor**:  Vassilios Tzerpos
 __Description__
-The Encyclopedia of Life (EOL) is a free, online collaborative
+Distributed Denial of Service (DDoS) attacks are recognized as one
-encyclopedia intended to document all of the 1.8 million living species
+of the most serious threats to today's Internet due to the relative
-known to science. It is compiled from existing databases and from
+simplicity of their execution and their ability to severely degrade
-contributions by experts and non-experts throughout the world.
+the quality at which Web-based services are offered to the end users.
+An especially challenging form of DDoS attacks are the so-called
+Application-Layer DDoS attacks. Namely:
-The ultimate criteria of success for a web application is the user
+  - In Application-Layer DDoS attacks, the attackers utilize a flood of legitimate-looking Layer-7 network sessions (i.e., sessions that are generally hard to detect and/or filter out by a firewall or an IDS system)
-experience. EOL is no exception. This project involves creating a
+  - Increasingly, these sessions comprise HTML requests generated by a cleverly programmed crawler that executes a semi-random walk over the web site links, thereby attempting to appear as a legitimate human visitor.
-framework for describing how visitors are supposed to interact with
-Encyclopedia of Life. It automatically checks that a new version the EOL
-code is functioning as expected. It is very beneficial for open source
-projects to be released often, and automated tests decrease the cost of
-releases dramatically and ensure the integrity of the data, code, and
-visual representation. The tests are run either in production or a
-duplicate of the production environment. The acceptance testing
-framework can be extended to check any website, not only EOL, by
-changing the configuration and defining new test suites. It also enables
-testing a web application with different browsers and operating systems.
-Acceptance testing is an import skill for aspiring developers. By
+The goal of this project is to investigate the state of the art in
-participating in this project you will learn the inner mechanics of
+malicious web crawler design. In particular, the project will look
-operating a browser automatically using scripts, emulating a real
+into the challenges of designing a smart-DDoS-crawler from the
-person's actions. This project will provide experience using Selenium (a
+attacker point of view - one of these challenges being the estimation
-leading open source acceptance testing tool), XML, XPATH, CSS selectors,
+of web-page popularity assuming no a priori access to the web-logs
-and Behavior Driven Development.
+of the victim web-site.
-More information: [[http://www.eol.org]]