Systems Research Overview

Parallelization of computations is a fundamental technique to accelerate software applications. With commercial success of massively parallel machines, research in this area has advanced from  analytical modeling of system performance, combined with simulation for validation of models, to actual design and implementation. Our most  recent success story is a large scale effort for parallelization of elliptic curve cryptographic (ECC) arithmetic, and pairing algorithms. The objective is to accelerate ECC and pairing based privacy preserving protocols so that they may be applied to real world applications. It achieves the highest computing speed in the literature. For details please following this link.

 Pattern Detection is at heart of pervasive Informatics on how to detect, organize and retreive zettabyte scales of information produced in the past two decades.  This information is often unstructured or structured in such a way that reality of the physical world the information represents is not immediately apparent. We have developed algorithms for detection of unknown (without signature) patterns, and anomalous traffic flows that can be used for an enterprise gateway. In a related line of efforts, we have developed complex pattern matching algorithms based on the notion of Character set constraint repetition (CCR). We have developed a MIN-MAX algorithm which solved two open problems: matching ambiguity and support of overlapped matching. The underlying matching model was expanded to matching of music melody encoded in MIDI format. We continue the development of powerful algorithms in string processing and information retrieval and apply these to workflow (business process) mining and geoinformatics.

Cyber-Physical Security  addresses the security issues of a system cyber-physical integrated system. Deep embedment of digital technology into nearly every corner of modern society introduces strong correlation between physical and cyber activities.   Through statistical inference of packet flows, a community interconnected through private channels on a network may unveil their (cyber and physical) organization and (cyber and physical) activities. We have both static and dynamic modeling efforts in this area.

For static modeling,  our on-going work is behavior modeling of adversary, who utilize the environment factors to  inflict low level conflicts in order to advance their interests.  We mainly focus on profit driven, risk averse type behaviors. By transforming these factors into an algorithm, the target selection  process may be formulated into an optimization problem. We aim to automate grading risk levels of geo-locations with respect to a specific type of conflict.  An ongoing project sponsored by the ONR develops such algorithms for grading of locations which may bear higher risk for IED attacks.  An extension of the concept is applied to screening of potential locations for unauthorized cultivation in public lands.

For dynamic analysis, our on-going work is to correlate RF/cyber signals that may be associated with a community’s activities. For a friendly community, our objective is assessment of team performance in a mission.   By identifying the relationship between electronic commands and the cohorts’ actions in the cyber and physical spaces, one can establish the normality of team performance, and also  mining of misbehaviors of team members.  For modeling of an adversary community’s activities, our objective  is situational awareness of RF signals which may be associated with their activities. A case study based on disturbance triggered audio surveillance can be found here. We have made substantial investment on software defined radio (SDR), Open BTS for analysis of GSM and related signals to interpret crowd behaviors. The research also touches on the integrity issues of GPS signals.

Cloud Confidence is the prerequisite for deployment of critical applications onto the cloud computing architectures. The trending emergence of cloud computing has shifted the focus of computing power from the PC to a distributed environment of utility computing, harkening back to the days of the mainframe. However, unlike mainframes, cloud technologies are not locally owned and maintained, leading to a slew of concerns for accountability, authenticity and anonymity. Therefore, we have developed novel cryptographic protocols and progressive system architectures for healthcare, social networking and security to propel Cloud Confidence.

In recent years, technology and information has become ubiquitous making humans and systems increasingly dependent on computing for real time operations. To handle critical industrial and mission oriented tasks, Cyber Physical Systems (CPS) are deployed in a pervasive manner in a variety of settings. CPS depend on high performance and integrity guarantees for automation, but not all environments currently allow for such assurance. In Tactical CPS we seek to integrate new technologies (software defined radio, chip scale atomic clocks and high performance/reconfigurable hardware), make provisions for advanced mobility and apply computing power to new areas in behavior modeling.

 

Contributing Members

  • Mike George
  • Hao Wang
  • John Pecarina
  • Casey Gonzales
  • Christopher Bodden
  • Joey Bartley

 

Leave a Reply