Runtime Optimal Semantic Gap-Filling System Security Monitoring via Offline Automated Executable Profile Generation and Dynamic Sensor Deployment

Saman Zonouz received a 3 year, $572K grant from ONR as a sole PI for the project "Runtime Optimal Semantic Gap-Filling System Security Monitoring via Offline Automated Executable Profile Generation and Dynamic Sensor Deployment".

The abstract follows.

Abstract: Secure operation of next generation critical cyber infrastructures requires not only intrusion preventative security hardening solutions, e.g., stack canaries, to prevent attackers from intruding the critical computer systems but also online optimal system security monitoring to provide semantic security status reports about low-level malicious activities within the systems. The objective of this research is to develop the foundations and working practical algorithms to provide adaptive and semantic gap-filling system security monitoring capabilities in complex and critical computing assets. The proposed automated self-aware and scalable cyber security solution will enable computing systems to survive cyber attack scenarios with minimum manual intervention and will provide the security admins with meaningful and concise reports about security incidents in real-time.

Active data screening for efficient feature learning

Waheed Bajwa and Anand Sarwate received a one year $160,000 grant from NSF CIF for the project entitled "Active data screening for efficient feature learning".

The abstract is given below.
 

Active data screening for efficient feature learning

Advances in sensing and data acquisition technologies make it easy to generate vast quantities of data that must be stored, communicated, processed, and understood. This data may be of variable quality and the nature of the data may vary over time -- this variability can cause difficulties for existing approaches to efficiently represent the data. Current methods use economical representations of the data in terms of a smaller number of properties, or features, of the raw data. Standard feature representations such as Fourier and wavelet representations may not be efficient at representing the data from these new acquisition technologies. One paradigm to overcome this mismatch is the data-driven approach, in which an algorithm processes the data to learn a novel and efficient feature representation for the given data. While these are more useful, such approaches may not scale well to massive data sets.

This work designs new methods for data-driven feature learning that are scalable and robust to noisy, time-varying data.

It proposes an "active screening" approach to learning new features; the data processing algorithm uses a low-complexity criterion to screen for useful and informative points for feature learning. Advantages of active screening include a reduction in the computational and storage overhead as well as the ability to reject outliers or other spurious and misleading data. The investigators develop active screening methods for consistent estimation under generative models for the data, analyze the tradeoff between representation and classification in active screening for discriminative dictionary learning, and extend the active screening analysis to distributed settings for distributed dictionary learning. They investigate the promise of these methods on two large-scale electrocardiography (ECG) datasets of 170+ patients. This work combines ideas from statistics (feature screening) and machine learning (active learning and selective sampling) to design efficient representations of complex signals from massive data sets and may inform the design of new data acquisition technologies by incorporating screening ideas into the technologies themselves.sarwate-activescreening.jpg

            Illustration of the benefits of active screening: selecting useful samples leads to better signal representations.

Spatiotemporally Varying Channel Map Estimation and Tracking in Wireless Networks

A. Petropulu and W. Trappe received a $500K 3-year CIF NSF grant for the project "Spatiotemporally Varying Channel Map Estimation and Tracking in Wireless Networks".   ECE PhD student D. Kalogerias was senior personnel on the proposal.

The abstract is shown below.

Spatiotemporally Varying Channel Map Estimation and Tracking in Wireless Networks

The availability of wireless channel maps can greatly improve the performance and reliability of wireless networks. In addition to traditional applications which depend on channel information, channel maps can be valuable in emerging applications such as communication-aware motion and path planning, network routing, connectivity maintenance and dynamic coverage, which will support improved wireless performance. In a realistic setting, the statistics of the wireless medium change dynamically in time and space. This research develops theory and algorithms for building wireless channel maps over a geographical area based on channel measurements obtained by the network nodes.

The descriptive statistics of the channel, referred to here as the channel state, are modeled as discrete time stochastic processes, evolving in time or space according to a fully or partially known statistical model. The channel state encompasses the path-loss exponent, the shadowing power and the correlation distance, and is hidden from the network nodes; the nodes can only observe their respective channel realizations. This project develops a novel framework for dynamic spatiotemporal estimation / tracking / prediction of both the channel state and the channel magnitude, in complex, nonlinearly evolving, time varying and possibly nonstationary environments. The estimation problem is approached through the rich theory of nonlinear filtering and stochastic control. Several issues are studied, including (1) Decentralized channel tracking & spatiotemporal channel prediction, (2) Event triggered sampling for efficient channel sampling, (3) Structured stochastic models for nonstationary channels. The project has an experimental component, which informs the analytical models and is also used to test/evaluate the developed methods. The project engages graduate and undergraduate students in a range of theoretical subjects and also measurements performed on WINLAB’s communications testbed.

Signal Processing--Optics Co-Design for In Vivo Optical Biopsy

Waheed Bajwa ECE and Mark Pierce BME received a grant from the NSF Engineering Directorate (ECCS Division/CCSS Program $359,986 for 3 years) for a multidisciplinary project, entitled "Signal Processing--Optics Co-Design for In Vivo Optical Biopsy."

Abstract:

For cancer and many chronic conditions, detecting early stage disease is the most critical factor in successfully curing patients and improving long-term survival rates. Current clinical practice involves taking biopsy samples from suspicious sites, followed by tissue processing and microscopic examination for abnormalities. This is a low yield, expensive, painful, and slow process. This project develops a new approach for microscopic imaging of living cells and tissues within the body in real time, which will improve the ability of physicians to detect early stage disease. The developed approach will also greatly increase the number of diagnostically useful biopsies being collected, improve the accuracy of margin identification during surgical resection, and permit non-invasive monitoring of post-surgical sites for recurrence. Lowering costs associated with unnecessary biopsies, multiple clinic visits, and repeat surgeries due to undetected residual disease will also positively impact the economics of healthcare delivery in the US.

The technical focus of this project is on the design of a fiber-optic probe that will enable examination of tissue for signs of disease in real-time, non-invasively, at the level of traditional pathology. This involves breaking conventional resolution limitations in fiber-optic imaging to deliver a real-time "optical biopsy." This is accomplished through integration of mathematical concepts from the compressed sensing field with hardware design and engineering for the clinical setting. The main goals of this project in this regard are (1) to design and engineer two candidate hardware architectures for optical biopsy, (2) to design, analyze and optimize the computational algorithms required to generate high-resolution images from these specific architectures, (3) to design and train automated feature recognition algorithms to assist the physician in interpreting optical biopsy images in real-time, and (4) to complete benchmark validation of the system using calibrated test targets and biological phantoms.

The intellectual merit of this project stems from the tight integration between fiber-optic-based endomicroscopy and signal processing theory and algorithms. This project does not simply apply image analysis algorithms to previously-acquired data in post-processing. Instead, the optical hardware components and signal processing elements are co-designed to enable imaging at 2-4 times higher spatial resolution than possible from developing the hardware or software alone.

This project will thus advance the medical imaging field by developing new signal processing techniques to increase resolution and field-of-view that do not require additional breakthroughs in microfabrication methods. This project will also advance the signal processing field by introducing new algorithms to solve missing data problems, address ill-posed inverse problems, and implement compressive sensing theory, all at previously unexplored length scale. However, the most transformative aspect of this work is that it has the potential to change clinical practices from reliance on 100 year old pathology practices to real-time information on tissue status at the patient's bedside.

Trustworthy and Adaptive Intrusion Tolerance Capabilities in Cyber-Physical Critical Infrastructures

Saman Zonouz received a 2015 NSF CAREER Award for the project "Trustworthy and Adaptive Intrusion Tolerance Capabilities in Cyber-Physical Critical Infrastructures". The grant has been awarded by the CNS division of CISE directorate. The total award is $508K and its duration is five years.

Developing an Application for Assessing Respondent Experiences of Their Surroundings in Real Time

Janne Lindqvist has received a new NSF award from the Sociology program with collaborators from Rutgers Sociology. The title of the project is "Developing an Application for Assessing Respondent Experiences of Their Surroundings in Real Time” and the award is for $51,847.

The abstract of the award is given below.

"Developing an Application for Assessing Respondent Experiences of Their Surroundings in Real Time”

The main objective of the project is to develop a robust method for measuring and understanding inequality in neighborhood experiences. The key thrust here is that we will implement and evaluate an app for smartphones to collect live information on people’s movement and neighborhood experiences as they happen. We will assess the feasibility of obtaining immediate explicit and implicit (nonconscious) measures of experiences in a place using the smartphone app. We will use these data to develop measures of mobile segregation and indicators of how people, in the moment, perceive their surroundings and opportunities in the places they go.

Obtaining data on how people explicitly and implicitly experience the places they visit, and how those experiences encourage or discourage them from going to certain locales, can help inform policy interventions that incentivize contact across demographic groups. The open source smartphone app developed in the course of this research will be shared with other researchers to advance knowledge on the connections between social, spatial, and cognitive processes and should have broad application beyond the specific study of social segregation in daily life.

A Software Defined Framework for Opportunistic Networking and Spectrum Management

ECE Prof. Narayan Mandayam and Ivan Seskar of WINLAB have received a grant from ONR entitled "A Software Defined Framework for Opportunistic Networking and Spectrum Management".

This is a $ 1M grant for 3 years (joint with Naval Research Laboratories, Rutgers share of funding is $790K).

The abstract is given below.

Abstract: Opportunistic transmission using non-contiguous chunks of spectrum is of emerging interest not only in tactical networking scenarios but also due to the increasing push towards coexistence and spectrum sharing between DoD and commercial systems. The last decade has seen the advent of software defined radio (SDR) based cognitive radios that have the ability to recognize signals received by them and adjust their own transmission frequencies, waveforms and protocols. In a network setting, along with cooperative techniques they hold the promise of promoting opportunistic operation that is required in tactical wireless networks and can provide performance gains using approaches such as collaborative signal processing, cooperative coding, relaying and forwarding. More recently, the paradigm of software defined networking networking (SDN) has emerged as a means to flexibly engineer networks by decoupling the functionality of the control plane and the data plane. In this project, we propose to combine SDR and SDN technologies to develop a comprehensive software defined framework for opportunistic networking and spectrum management. Such a framework will allow radios (and systems) in a tactical environment to dynamically and opportunistically transmit in non-contiguous portions of spectrum, provide robustness against radio channel variations due to both electromagnetic propagation and interference, as well as adversarial conditions such as attacks. The mechanisms developed under the above framework will fundamentally rely on the ability of SDR based cognitive radios in the network to execute non-contiguous multicarrier modulation and use non-contiguous Orthogonal Frequency Division Multiple Access (NC-OFDMA) where by non-contiguous subcarriers can be flexibly assigned across nodes. An accompanying SDN based control plane architecture will be designed for implementing such opportunistic spectrum sharing mechanisms in networks such as those relevant to the Navy.

Directional Network Waveform

Wade Trappe (ECE/WINLAB) and Ivan Seskar (WINLAB) have won a contract from the Army (funded in collaboration with DSCI) titled Directional Network Waveform. The funding for the contract is $112,149 with a performance period of 5/21/15 to 9/30/15.

The abstract is shown below.

Congratulations to Wade and Ivan on this achievement!

Abstract:

Directional Network Waveform

Traditional ad hoc networking protocols enable a local node to gather information about itself; however, this does not transition well into understanding the network holistically. Tactical networks require the need to operate effectively under the presence of disruption, attack, congestion, or local failures. By utilizing a distributed control plane that shares node state information in conjunction with a hierarchical and distributed cognitive engine that can tune cross-layer network protocols and algorithms, it is possible to move information across the network occurring at orders of magnitude greater than current approaches, with better survivability and with less interference using directional antennas. This research effort will define and develop a directional networking software architecture based on provisioning a separate control plane that supports cognitive distributed control for adaptively managing the resources of mobile ad-hoc networks equipped with directional antennas so as to provide robust and high-levels of communication performance for tactical mission operations. Development efforts will focus on FPGA-based implementation of communication algorithms needed in support of such a directional networking waveform.

Pages

Subscribe to Rutgers University, Electrical & Computer Engineering RSS