CS Students Win at
University Research Forum
Posted on 2015-04-01
Ph.D. candidates Abhijit Nag and Daqing Yun received 1st and 2nd place, respectively, in the Math/Computer Science category of the poster presentation at the U of M's 27th Annual Student Research Forum held on March 30.
Abhijit's work under Prof. Dipankar Dasgupta is titled "Design and Implementation of an Adaptive Multi-factor Authentication(A-MFA) Framework." Multi-factor authentication is currently an urgent need to facilitate continuous protection of computing devices and other critical online services from an unauthorized access. Many authentication mechanisms with varying degrees of detection rate of users and portability are available for different types of computing devices. Hence, the adoption of a multi-factor strategy in authentication process provides a flexible solution to the users as well as protecting their identity. This research focuses on designing and developing a trustworthy model for the authentication factors (modalities and their features) based on different types of operating devices and connected media. Moreover, a novel probabilistic approach capable of adaptive and non-repetitive selection of different authentication factors (for a set of devices and media) is also proposed in the research. This selection approach also considers their (different factors) computational complexity, trustworthy values, performance in authentication selection, and previous use of the selected modalities in the same settings. The trustworthy values derived from the proposed trustworthy model are the guiding factors for adaptive selection of the authentication factors with the consideration of the effects of surroundings (light, noise, etc.) on different categories of devices and media. Empirical studies demonstrate that the proposed selection approach performs better than other selection strategies in different environmental settings.
Daqing's work under Prof. Chase Wu is titled "Technologies and Tools for Synthesis of Source-to-Sink High-performance Flows: Transport Profile Generation." The transfer of big data is increasingly supported by dedicated channels in high-performance networks. Transport protocols play a critical role in maximizing the link utilization of such high-speed long-haul connections. The optimal operational zones of such recently proposed transport protocols such as UDT are affected by many factors and their default values do not always yield the best performance. In addition, scientific application users typically do not have the knowledge to choose which transport protocol to use and which parameter value to set. This work proposes a Transport Profile Generator (TPG) to characterize and enhance the end-to-end throughput performance of transport protocols. TPG automates the tuning of various transport-related parameters including socket options and protocol-specific configurations, and supports multiple data streams and multiple NIC-to-NIC connections. TPG optimizes CPU affinity settings of sending/receiving threads in multi-core system by exploring the internal connectivity/topology of various hardware components (cores, sockets, NICs, etc.) through interfacing with the Host Profiler provided by ORNL. TPG also provides profiling API functions to be called by other tools such as XDD to improve the data movement performance. To instantiate the design of TPG, this research uses UDT and TCP as examples in the implementation and conducts extensive experiments of big data transfer over high-speed network channels to illustrate how existing transport protocols benefit from TPG in optimizing the throughput performance.