|
17 |
|
|
Description A substantial amount of research effort has been devoted over the past decade on developing network intrusion detection (NID) systems, and today we have a number of sophisticated NID systems available commercially. Nonetheless, malicious attacks do occur and the results of these attacks are often serious. Current NID systems are able to detect these attacks most of the times. However, remedial actions are still taken by system administrators and involve shutting down the network temporarily. The delay in response is often enough for intruders to launch attacks on other networks from the compromised networks. This delay may not always be acceptable for critical networks. The current project addresses the problem of rapid response to intrusion alerts so as to isolate and contain intrusions. We plan to develop a proof-of-concept for a computer network that is able to defend itself on its own against malicious attacks. This network is able to isolate any compromised subnet rapidly and automatically, thus protecting the remainder of the network from further damage. It is able to confine the attack to the compromised subnet only, thus preventing the use of the compromised subnet as a launching pad for further attacks. Our goal is to minimize human intervention needed to achieve all this. The project has two major components:
(1) Intrusion Detection
(2) Intrusion Isolation and Confinement |
|
|
18 |
|
|
Description Waveform modeling of signals through a communication channel is very important for determining system level performance. One may want to determine the effect that a nonlinear device have on system performance. The communication channel may be exhibit temporal and/or spatial correlations and one may wish to ascertain its effect on different coding schemes. I am proposing to develop a model of a system and run monte carlo simulations to determine performance based on the applicable scenario. |
|
|
19 |
|
|
Description Many applications-in numerical simulation, physical modeling, multimedia, data mining, multimedia, and signal and image processing- spend a significant part of their time in loops. Often these parts of the program are not optimized to extract the best performance. We have developed considerable expertise on a number of program transformations that show considerable promise for optimizing the performance. For some loop kernels that occur in biological sequence comparison and gene sequencing, we have achieved up to 300% improvement We believe that our techniques can be used in many different contexts, for different (computation intensive) applications and for different target architectures, such as modern multiprocessors with a sophisticated memory hierarchy and deep pipelines and vector and parallel machines and supercomputers. They can also be useful for directly "compiling" the application to silicon (designing application specific circuits, in the context of high performance embedded and consumer electronic systems (games, cameras, image processors, printers, etc.) Furthermore, it is possible to optimize for different performance criteria (speed, power and/or energy consumption, memory, silicon area, etc.) |
|
|
20 |
|
|
Description Testing software that runs on distributed, heterogeneous platforms presents even more difficult challenges than testing single programs. Our research is aimed at enabling software testers to apply rigorous testing methods for integration and system testing. We are developing techniques that will aid in the management (both monitoring and control) of test execution in the distributed setting. We are adapting existing fault injection techniques for assessing the fault tolerance capabilities of distributed applications. We are also developing and evaluating means for measuring the quality of testing performed. We are experimenting with test adequacy criteria and evaluating their effectiveness in measuring the quality of testing. |
|