Department of Computer Science

PURDUE UNIVERSITY

 

CS Participants in Undergraduate Research Day

Sponsored by College of Science (CoS)

 
Student(s)
Research Advisor
Project Title

Brett Bavar

Dr. Cristina Nita-Rotaru
On the Security of Beacon Vector Routing
Kit Chan Dr. David S. Ebert Analysis of Emergency Planning and Response with 3D Modeling on
Mobile Device
Yu-Hsuan Rex Cheng
Dr Daisuke Kihara Benchmark of new amino acid similarity matrices on recognition of
distant protein structures
Tom Scogland Dr Cristina Nita-Rotaru Experiments on Multi-tree Overlay Networks
Ryan Stutsman Dr Sonia Fahmy Translation-based Steganography

 

 

On the Security of Beacon Vector Routing
Brett Bavar, Computer Science
Research Mentor: Cristina Nita-Rotaru, Computer Science
(back to top)

With recent proposals of more sophisticated applications for wireless sensornets, new challenges have arisen in applying very limited sensor mote hardware to more difficult tasks. Although Beacon Vector Routing (BVR) is well-suited to these unique challenges, it does not provide any security against routing attacks. First of all, BVR is susceptible to multiple types of well-known sensornet routing attacks. In addition, I have identified a BVR-specific attack, which I call beacon takeover. Beacon takeover could be used by an attacker to gain a dangerous amount of control over routing for the network.


Analysis of Emergency Planning and Response with 3D Modeling on Mobile Device
Kit Chan, Computer Science
Research Mentor : David S. Ebert, Electrical and Computer Engineering
(back to top)

We are creating software to display 3D models on mobile devices for use in emergency response. By generating a 3D model of the environment to visualize the
emerging situation inside the building and record the complete process for later analysis. In addition, different sensors can increase the functionality of the system
to provide more accurate and practical data. With the mobility of the mobile computing system, it provides better adaptability to different kinds of physical and
computational environments. This research will allow better planning for emergency situations.


Benchmark of new amino acid similarity matrices on recognition of distant protein structures
Yu-Hsuan Rex Cheng, Computer and Biological Sciences
Research Mentor : Daisuke Kihara, Computer and Biological Sciences
(back to top)

An important initial step for protein structure prediction is to find a template structure in a database which fits to a target protein sequence. A template structure
can be easily found by BLAST (Basic Local Alignment Search Tool) with a conventional scoring matrix, such as BLOSUM62, if the template has a sufficient
sequence similarity to the target sequence. However, distantly related template structure is not visible by BLOSUM62, because it is designed for finding
homologous sequences. In our previous study, we have developed novel amino acid scoring matrices that are proven to be more useful in aligning pairs of distantly
related protein sequences than existing matrices. Here we further test the newly developed scoring matrices in searching distantly related protein structures with a
database by using BLAST. For this benchmark, we have used the Lindhal and Elofsson database which classifies alignments in the fold, superfamily and family
levels. This report describes the results and accuracy of the new developed matrix, and further more, describes the similarities and differences in the results comparing to the standard alignment scoring matrix.


Experiments on Multi-tree Overlay Networks
Tom Scogland, Computer Sciences
Research Mentor : Cristina Nita-Rotaru, Computer Sciences
(back to top)

Overlay broadcast systems provide broadcast and or multicast services at the application level. These services are needed for networks which don't implement
their own broadcast or multicast services, like IP. Most overlay broadcast systems use a tree, rooted at the sender, to disseminate the information to all the targets. More recent designs however use a multi-tree approach to avoid imposing all the cost of sending the message on the inner nodes, and allowing for a more fair bandwidth distribution. Security of these systems is a potential problem for several different reasons. First off, most of the systems in use today do not have security in mind in the design phase, and thus may well present vulnerabilities. Since they do not take that into account, it is important to determine the threats possible if a machine on the overlay network is compromised, or even perhaps owned by an attacker. Some common problems that might be caused using networks like these are denial of service attacks, data corruption or replacement on the network, and corruption of the overlay network itself. During the course of this project the intent is to study the viability of some forms of attack against the SplitStream overlay multicast system, as well as testing its resilience and reliability under the influence of the attacks.


 

Translation-based Steganography
Ryan Stutsman, Computer Science
Research Mentors: Sonia Fahmy, Computer Sciences
(back to top)

This work investigates the possibilities of steganographically embedding information in the "noise" created by automatic translation of natural language
documents. Because natural language translation inherently creates plenty of room for variation, it is ideal for steganographic applications. Also, because there are
frequent errors in legitimate automatic text translations, additional errors inserted by an information hiding mechanism are plausibly undetectable and would appear
to be part of the normal noise associated with translation. Significantly, it should be extremely difficult for an adversary to determine if inaccuracies in the translation
are caused by the use of steganography or by deficiencies of the translation software.

 


Complete List of Abstracts/Participants

Video of the Event