 |
| ||||
|
|
| ||||
| Network Monitoring | Data stream algorithms for monitoring TCP/IP networks for intrusion detection and traffic behavior characterization: Due to the ever-increasing size of the internet and the rapidly growing number of web applications, the volume of data exchanged over the internet today is enormous. Network monitoring applications constantly monitor computer networks for identification of slow or failing components, detection of possible intrusions like worms or Denial of Sevice attacks and characterization of usual network traffic behavior. A network flow is usually defined by the attributes like source IP address, destination IP address,
source port, destination port and type of service. Network monitoring applications (e.g., products that implement the Cisco NetFlow protocol) collect these flow data by listening to network interfaces; and analyze these data to detect any potential threat to the network, by searching for patterns that are characteristic of typical exploits (worms or DoS attacks). Our goal is to design and implement efficient algorithms that work on massive volume of real network flow data, but require small space to mine different patterns (that characterize typical exploits) from this data.
| ||||
| Distributed Data Mining | Computing Frequent Elements using Gossip: Due to the large scale of P2P and sensor networks, the values of
aggregate functions over the data in the whole network are often more important than individual data at nodes. For example, in a P2P file-sharing network (like Gnutella or Napster), an administrator may be typically interested in identifying the most frequently accessed items over the whole network. Any reliance on central coordination limits the system’s scalability, therefore gossip-based protocols are emerging as an important communication paradigm. In gossip-based protocols, each node contacts one or a few nodes in each round (usually chosen at random), and exchanges information with these nodes.
This mechanism leads to high fault-tolerance and self-stabilization. Our work is to design efficient distributed algorithms for identifying the frequently occurring data elements with probabilistic guarantees on accuracy,
where the nodes would exchange small-sized "sketches" or synopses of their individual data sets through gossip algorithms.
| ||||
| Sensor Networks | Development of Distributed, Lightweight, Secure Hierarchical Directories for Tracking Mobile Objects in Sensor Networks:   Most practical techniques for locating remote objects in a
distributed system suffer from problems of scalability and locality of
reference. We are implementing the Arrow distributed directory protocol, a
scalable and local mechanism for ensuring mutually exclusive access to
mobile objects. This directory has communication complexity optimal
within a factor of (1+MST-stretch(G))/2, where MST-stretch(G) is the
minimum spanning tree stretch of the underlying network. In Arrow protocol, local change
in the object’s position does not result in a global change in the network. This has been
deployed on a fixed spanning-tree-based network of MICA2 motes, where the presence of the object is detected
by measuring the amount of ambient light using MTS300 photo sensors, and the node detecting the object
transmits this information to its neighbours so that the pointers are updated appropriately.
| ||||
| Business Applications | DADS (DUoS and Associated Distribution Services):
The goal of the DADS program was to meet the regulatory requirement to separate
electricity distribution and supply businesses of United Utilities and develop
an IT solution that will cater to the pure distribution business. The application developed
by TCS provided a common web-based interface across all the distribution functions and it
replaced a number of different legacy applications that were used for distribution billing previously.
Roles played:
Trailblazer (Acuity): This web-based application developed for McGraw-Hill Digital Learning enhanced the functionality of an existing application named TrailBlazer with creating test, exercise and assignments for students; scoring the students' response and generating predictive, diagnostic and summary reports. The system also monitors the structure of state curriculum and tracks the progress of a student in class. Roles played:
| ||||
|
|
| ||||
 |
Bibudh Lahiri Department of Electrical & Computer Engineering Iowa State University 3125 Coover Hall, Ames, Iowa | ||||
