Sudeep Sarkar - US grants

The evolution of geometric and photometric attributes of 2D organizational structures can not only provide evidence for the underlying structure and motion but also help infer complex motions. This research will develop a computational theory to extract structural and motion information from long image sequences by exploiting perceptual organizational principles. Simple motions will be grouped to infer about complex motion. An efficient computational framework will be developed to monitor the evolution of the organizational structures. Detection of the 2-D organizations isbased on graph theory and voting methods, and a modification of Bayesian networks. These 2D organizations are then tracked through the image sequence to generate traces of their photometric and geometric properties. This research will open up possibilities for the development of new tools for motion understanding using organizational principles.

CDA-9724422 Sarkar, Sudeep University of South Florida Acquisition of a Cyberware 3D Scanner to Facilitate State of Art Research in Computer Vision and Graphics The University of South Florida is acquiring range scanner and graphics workstations to support research in computer vision and computer graphics. Specific research projects include performance characterization for range segmentation algorithms, color-texture analysis for range and color images, non-rigid motion estimates from range images, and algorithms for processing digitized data using non-uniform rational B-splines. A number of graduate and undergraduate students will be using the proposed equipment in research projects. The University is also considering using the equipment in courses such as Computer Vision, Image Processing, Pattern Recognition, and Computer Graphics.

9729904 Bowyer, Kevin W. University of South Florida CISE Research Instrumentation: Acquisition of a Compute Server for Image Analysis Research that Emphasizes Empirical Performance Characterization This research instrumentation grant contributes to the purchase of a compute server for the lab research network at USF with four 250 Mhz UltraSparc Processors, 1GB main memory, 35 GB of formatted disk pace, and a tape drive, which will enable the following projects:- Automated Performance Evaluation of Edge Detectors,- Nonrigid Motion and Structure Recovery from 2D Views, - Performance Evaluation of Perceptual Organization Modules, and - Automatic Tumor Volume Extraction from Brain MRI. This project deals with four image analysis tasks: performance analysis of edge detector algorithms, algorithms for analysis of non-rigid motion in image sequences, the development and analysis of perceptual grouping algorithms, and the development of clustering-based algorithms for the segmentation of magnetic resonance images. These tasks are all computationally expensive. In addition, the studied empirical characterization of the performance of the algorithms is emphasized. This style of research presents enormous computational demands, and it is not feasible to conduct this style of research with the typical workstation that an individual user would have on their desk. To make it practical to carry out this research, USL will acquire a compute server that will provide much greater computational power. This compute server will be shared among the research projects. Each of the projects is innovative and will lead to high-quality publications. In addition, the data sets and empirical methods that are created in these projects will have substantial impact on future research.

Abstract

IIS-9907141
Sarkar, Sudeep
University of South Florida
$73,407 - 12 mos.

The Role of Learning in Perceptual Organization of Complex Images

This is the first year funding of a three year continuing award.
Perceptual organization is the ability to efficiently group low-level
primitives, such as pixels or edge segments or regions, so as to form
object hypotheses with minimal domain knowledge. This process is an
essential link between the noisy low-level segmentation processes and
the sophisticated, computationally expensive, object recognition
processes. This work will address three major scientific issues
related to perceptual organization in computer vision: (i) How to
efficiently form large groups of extended low-level primitives by
assembling small groups of features (grouplets)? The concept of graph
spectra is an ideal mechanism for this purpose. (ii) How to encode
the dependency of grouping parameters on image statistics? Joint
probability distributions encoded as Bayesian networks offer an
efficient abstraction. (iii) What is the role of learning in
perceptual organization? The specific roles of learning explored would
be threefold: (a) learning the image to grouping parameter mapping,
(b) learning the forms of the groups, and the associated hierarchy,
from object models, and (c) learning the primitive type, such as
corners, textured-regions, or edges, that is appropriate for a domain.

Computer Science (31)

This proof-of-concept project is developing educational materials to integrate research progress in image related computation (computer vision, image processing and image analysis) into the undergraduate computer science and engineering core curriculum. Recent years have experienced an explosion of image gathering modalities along with a resultant increase in the demand for expertise in image related computation. Considerable progress has also been made in fields such as image processing, image analysis, and computer vision. It is no longer sufficient to address this need for image expertise through one upper level elective course in computer vision. The need to incorporate image related knowledge units in core undergraduate courses such as data structures, introductory programming courses, automata theory, computer ethics, databases, networks, coding theory, and computer algorithms, preferably without major change in the content of these courses is essential.

The materials for these knowledge units are being developed so that they can be used by instructors who are not necessarily computer vision specialists. An added advantage of using image related knowledge units in various courses is that, because of their inherent visual nature and large sizes, images offer an excellent medium for the better understanding of underlying core computer science concepts.

As deliverables, the project is developing specific image-related knowledge units and instructional support materials, such as class handouts, transparencies, lecture materials, software, and descriptions of active learning contexts. Dissemination is being accomplished through the development of a textbook and freely available website. The accumulated experience and evaluation results are being presented at major computer science education conferences, workshops, and through journal papers.

The objective of this research is to extend the knowledge of the single player stochastic decision problems, which have been studied for years, to develop new methodologies for multi-player games and test them on large-scale problems from the domain of e-commerce and supply chain management. Such a methodology, when developed and tested, will provide a much needed resource to the corporate world for examining business policies. The Internet revolution has brought about tremendous changes in the marketplace by tearing down the barriers of time and distance. The competition among the providers of goods and services for luring the customers has reached an epic height. Consider, for example, a homebuyer's request for a mortgage loan, which is now available to virtually every lending institution (e.g., a bank) in the world. All the banks (players, in generic game theory nomenclature) seeking to capture this customer are involved in a stochastic game, where they form their bids in anticipation of other players' actions. Based on the outcome of their bids, the players try to learn a strategy for the subsequent customers. In the above example, the game environment is highly stochastic and the game return is not necessarily of the zero-sum type.

The main purpose of modeling and examining such problems is to foresee the equilibrium point(s) of a game, the path of the game evolution which tells us about the game returns in finite time horizons, and also the time taken by a game to reach an equilibrium point. In today's volatile market situation, short-term game returns could be of much more immediate importance. Standard game theoretic analysis, when available, is geared toward characterizing the equilibrium points. Through simulation-based approaches to the study of stochastic games, as presented in this project, one can also examine the evolutionary paths a game can follow.

EIA-0130768
Sudeep Sarkar
University of South Florida

CISE Research Resources: A Compute-Intensive Sensor-Based Environment for Research in Computer Vision and Artificial Intelligence

Automated learning of grouping parameters for perceptual organization of complex images, modeling and reconstruction of elastic objects from image sequences, real-time matching of buyers and sellers for E-commerce, and learning models from extremely large databases, all require large data storage and a computing environment that supports exploring extremely large parameter spaces along with the ability to process huge quantities of data. A multiprocessor computing environment with substantial memory and disk storage is requested for high-performance computing associated with these four research projects in the general areas of computer vision and artificial intelligence. The compute server will increase the present capabilities by an order of magnitude.

In addition, image acquisition devices, including high-resolution color cameras, digital video cameras, stereo cameras, and laser range scanners are requested for gathering color, motion, and range data. The ability to acquire fast range images and motion sequences will enable the consideration of the problem of integrating motion and range into the perceptual organization process. Also, the ability to acquire fast and high-resolution range, with registered color, will facilitate development of physics-based non-rigid algorithms and models that incorporate true material properties, which have, heretofore, not been possible.

This award entails collaborative research between the University of South Florida and Rowan University to develop a methodology for three-dimensional quantitative characterization of particle morphology, as well as to examine the effect of particle shape and angularity on the mechanical response of geomaterials. New methods for three-dimensional characterization and imaging of particles are proposed. Recent innovations in geotechnical particle shape identification in two dimensions, supported by emerging technologies in digital image analysis, shape recognition, and object reconstruction will be adapted to map the three-dimensional shapes of particles. Quantitative measures for particle morphologies in three dimensions will also be introduced. A link will be established between 2D and 3D particle shapes through digital image analysis and pattern recognition. A digital library of particle shapes will be made available online to the scientific community. Methods and algorithms recently developed by the PIs to model angular particle shapes in 2D will be extended to 3D, and the particle geometries obtained using the 3D image analysis will be incorporated within a Discrete Element Method (DEM) code. The influence of particle shape on dilatancy and steady state shear strength will be evaluated through experimental measurements and DEM simulations. The advantages and limitations of the new techniques will be assessed accordingly. The proposed research will involve several undergraduate and graduate students at both institutions. In addition, teachers will be engaged in developing modular exercises for middle and high school students through institutional RETs. This alone will be greatly beneficial to the large fraction of the population.

Sign languages are complex, abstract linguistic systems, with their own grammars, and their "articulation" involve not just the hands, but the face, shoulder, and arms. In this project the PI and his team will push the state of the art in scalable automated American Sign Language (ASL) recognition formalisms. Presently, there are methods to recognize isolated signs and, to some extent, continuous signs in short sentences from a single signer, mostly using special equipment such as data gloves or magnetic markers or from visual input against plain background and special clothing. The PI seeks to achieve substantial advances in five areas:: (i) recognition against varied backgrounds and different clothing, (ii) use of non-manual aspects such as facial expression and head movement, (iii) recognition across signers, (iv) design of robust, scalable formalisms, and (v) development of large ASL data corpus that exercise variates such as viewpoint, background, time, and signer, to help benchmark progress. To these ends, the PI will (i) develop robust manual and non-manual (face) feature sets for ASL, (ii) construct formalisms to learn sign models from example sentences, (iii) investigate if elemental forms of signs (signemes) can be learned, (iv) employ Bayesian network based indexing schemes to limit the combinatorics of recognition, and (v) explore techniques to incorporate grammar and syntax information into the recognition process.

Broader Impacts: With the gradual shift to speech based I/O devices for human computer interaction, there is great danger that people who rely on sign languages for communication will be deprived access to state of the art technology unless there significant advances in automated recognition of sign language are achieved. Such advances will also enhance the quality of life of persons with disabilities, by facilitating interaction with the general populace in public situations, such as airports and grocery stores. The PI will identify at least one deaf graduate student or a student with communication disorder to participate in the project, to ensure the outcome is relevant and appropriate to the intended user community. The large ASL data corpus collected in this project will be distributed aggressively; this effort will continue to be supported beyond the project end date.

0829838 - EMT/Nano: Energy Minimization Computing using Field Coupled Nanomagnets--Modeling and Fabrication

Sudeep Sarkar, Sanjukta Bhanja

Field coupled computing is a radically different paradigm where electrical, magnetic or spin coupling among nano-devices are utilized for computation. By far, most advances has been in the nano-fabrication of nanomagnets, mostly driven by the need for denser memory and patterned magnetic storage media. This research will open up unconventional front in computing. Unlike other current works in nanocomputing that seeks to replicate traditional computing involving Boolean logic, this research is using magnetic field-based computing (MFC) to solve optimization problems in computer vision. In the long run, this research will help design computers that will be able to solve hard problems in automated object recognition in a computationally efficient manner.

The ground state of a nanomagnet collection minimizes the Hamiltonian that is governed by the pairwise dipolar interactions between the nanomagnets. The specific focus of this research is to harnesses this energy minimization aspect to solve quadratic energy minimization problems in computer vision. The investigators are developing computational method, based on multi-dimensional scaling, to decide upon the spatial arrangement of nanomagnets that matches a particular quadratic minimization problem. Each variable is represented by a nanomagnet and the distances between them are such that the dipolar interactions match the corresponding pair wise energy term in the optimization problem. The nanomagnets that participate in a specific computation are to be selected from a field of regularly placed nanomagnets. Some of the scientific questions being considered are. What would be the optimal geometry and appropriate material for these nanomagnets? How would one select the nanomagnets from an array of uniformly spaced nanomagnets to involve in computation? Can we demonstrate these functions by fabricating circuits?

Proposal #: 12-29561
PI(s): Dubey, Rajiv; Diamond, David M; Quillen, William; Reed, Kyle; Sarkar, Sudeep
Institution: University of South Florida
Title: MRI/Acq.: CAREN: Virtual Reality System for Collaborative Research in Assistive & Rehabilitation Technologies

Project Proposed:
This project, acquiring an instrument referred to as CAREN (Computer Assisted Rehabilitation ENvironment), aims to greatly facilitate ongoing interdisciplinary and inter- and intra-institutional research to analyze human mobility and function and to improve the quality of life of individuals with disabilities and older adults by increasing their independence and community reintegration. The instrument, a turnkey customizable 3D virtual reality system with applications in complementary research and training experience in the broad areas of rehabilitation engineering and science, also includes a cylindrical screen projection system, 12-camera real-time motion-capture system, a six degree-of-freedom motion base and a control software suite. The team includes about 20 PIs from various departments at the institution, which currently lacks a shared virtual reality simulation facility for large collaborative projects. The instrumentation, to be used for fundamental research in various disciplines, analysis and creation of innovative technology solutions that can be rapidly evaluated, tested, prototyped and commercialized, is expected to dramatically reduce the need for costly physical simulations and allow complex testing not otherwise possible. The following research projects will be supported by CAREN:
- Fundamental research: measure effects of sensory inputs (e.g., visual, auditory, vestibular, tactile);
- Early diagnosis related research: quantify behavioral indicators for early detection of disorders;
- Applied rehabilitation research: provide real-time feedback to the user, thus allowing rapid correction of therapeutic movements;
- Sports medicine research: optimize training techniques for peak performance and injury prevention; and
- Research on rehabilitation transfer to daily life: model environments that allow more real-life training for enhancing laboratory or clinic based rehabilitation.

Broader Impacts:
The impact should be felt both regionally and nationally. At the national level, the proposed system could have great implications for improving the quality of life of individuals with disabilities and older adults. At the local level, the VR equipment will be integrated into a number of courses and will engage K-12 students. It is expected that the engagement of students through the REU program can greatly impact undergraduate research. There is a substantial industrial partnership and potential for technology transfer.

It is natural that events of interest in observed scenes manifest themselves across multiple sensing modalities - vision, hearing, smell, etc. The remarkable perceptions of audio-video signals by natural systems, such as humans, also points to superiority of inferences drawn across modalities. It, therefore, seems natural to enhance performance of automated systems by using joint, cross-modal statistical inferences. However, the detection, organization, and understanding of cues and events in real-world scenarios are difficult tasks. This project seeks to develop a pattern-theoretic framework for achieving these goals. The main research items are: (1) development of mathematical quantities to represent audio and visual events and their spatiotemporal relations, (2) use domain-specific ontologies to impose semantic structure and to incorporate prior knowledge, and (3) derive algorithms for Bayesian inferences using efficient adaptations of Markov Chain Monte Carlo sampling.

The use of pattern theory allows bridging of gaps between raw signals and high-level, domain-dependent semantics, and helps discovers large groups of audio-visual events likely to represent the same underlying event. This effort combines ideas from perceptual organization in computer vision, computational analysis of auditory signals, pattern theory, and prior developments in ontological structures. The methods developed here are applicable to many scenarios that deploy audio and video sensors, including problems of audio annotations of videos, speaker tracking in teleconferencing, and separation of multiple objects in remote surveillance. Broader impact activities involve the development of teaching modules, innovation and entrepreneurial training of the students, and communication of the findings to the community.

This project establishes an I-Corps Site at the University of South Florida (USF). NSF Innovation Corps (I-Corps) Sites are NSF-funded entities established at universities whose purpose is to nurture and support multiple, local teams to transition their technology concepts into the marketplace. Sites provide infrastructure, advice, resources, networking opportunities, training and modest funding to enable groups to transition their work into the marketplace or into becoming I-Corps Team applicants. II-Corps Sites also strengthen innovation locally and regionally and contribute to the National Innovation Network of mentors, researchers, entrepreneurs and investors.

The site at USF institutionalizes the Lean LaunchPad training process at the Center for Entrepreneurship (CFE) and transforms processes in its Technology Transfer Office (TTO) to facilitate faculty and students to transfer STEM research into commercially viable ideas and prototypes. Faculty and students are recruited for this Site using pro-active processes such as: visits to classrooms, capstone design courses, research projects on campus, undergraduate and graduate research displays, individual contacts with faculty with STEM related research grants, and invention disclosures. The TTO also helps teams connect with business mentors by leveraging current contacts. They also are a resource for finding non-NSF funds for innovation-related activities. Faculty from CFE, representing the Colleges of Business, Engineering, Medicine and Sustainability, deliver the NSF training program in a one-week, intensive boot camp format.

Beyond the prototype stage, USF CONNECT, which is the business arm of USF and at the epicenter of Tampa Bay's innovation ecosystem, supports the I-Corps teams throughout the business life cycle, providing facilities, partners, and resources for business development and access to technologies, workforce programs, technology commercialization, critical research equipment, and incubator facilities. The I-Corps program results in an institutional impact by addressing a crucial gap in the USF innovation ecology in creating prototypes and enhancing design.

With an I-Corps site, the potential for increasing STEM technology transfer that feeds Tampa Bay's growing innovation ecosystem is high and contributes to transforming the region through high-technology startups. USF is an economic anchor for the Tampa Bay region and a node of the Florida High Tech Corridor that spans a 23-county region in Florida. Most of the state's medical device manufacturing, defense, aerospace, informational technology, and life sciences companies reside in this region.

USF's diverse student population are also targeted in the I-Corps Sites program. NSF/USF supports the Bridge to the Doctorate and Sloan Minority PhD programs that have enabled USF to be ranked as a top-10 producer of minority (African American/Black, Hispanic/Latino) engineering doctorates. The I-Corps site also nurtures innovation among students who are military veterans, as more than 2,100 veterans and their families are enrolled as students at USF.

The world of computing has entered the multi-core age. In addition to multi-core CPUs, co-processors containing thousands of computing cores in a single chip have become popular platforms for general-purpose computing. With the aggregated computing capabilities increasing at a steep rate, computing communities are still in an early stage in developing software systems, frameworks and applications to take full advantage of these new platforms. The co-existence of several different multi-core systems, including the Graphics Processing Units (GPUs), Intel?s Many Integrated Core (MIC) cards, and Accelerated Processing Units (APUs), further complicates the issue. This, on the other hand, provides opportunities for interesting research that spans different layers of the software stack. This infrastructure will support multiple, coordinated research projects that will develop frameworks and software systems enabling a new class of applications requiring high-performance computing capabilities.

The main goal of this project is to build a computer cluster with heterogeneous, massive parallel computing capabilities to accelerate existing research and enable ground-breaking new research that shares the same need for intensive computation at the University of South Florida (USF). This project brings together eight USF investigators with research projects in several core disciplines of computer science and engineering: big data management, scientific computing, system security, hardware design, data mining, computer vision and pattern recognition. Specifically, the requested cluster supports research in:
(1) design and optimization of a novel data stream management system architecture in a heterogeneous many-core hardware environment;
(2) coarse-grained molecular simulation approach that allows accurate simulation of large-scale atomistic systems;
(3) new system to deploy security policies that excel in both policy composition and runtime performance;
(4) efficient modeling and design of energy-efficient and secure hardware systems;
(5) automated interpretation of activities using pattern theory;
(6) fast large scale clustering; and
(7) pattern identification from biomedical image data.
The intellectual merit of this project derives from the innovations of the individual projects and from the potential cross-disciplinary ideas it can germinate in the future. The infrastructure is expected to facilitate collaboration and cross-pollination of algorithms, models, representations, and data sets across individual project areas, building a collaborative network across the investigators. Furthermore, the cluster is expected to impact over a dozen application domains via on-going and planned research projects among the investigators and their collaborators throughout the USF system.

Direct benefits to education and research will also be extended to the larger community through the applied aspects of projects, teaching and training. Project results and media content of the cluster will be showcased in the popular USF Engineering EXPO event, which seeks to educate and motivate K-12 students on math, science, engineering, and technology subjects.

The broader impact/commercial potential of this I-Corps project involves computer vision analysis of video, using both visual and auditory cues, to create descriptions of the content. The technology has a large variety of potential applications from law enforcement to surveillance to consumer applications. These include enabling the efficient storage and retrieval of large volumes of camera data. Smart surveillance systems can be enhanced with features that allows for summarization of daylong video footages as a list of security-relevant events. The technology can also allow automated organization of large collections of multimedia data.

This I-Corps project involves commercialization feasibility research for a computer vision technology for expressing video content in terms of natural language text and grammar, i.e. semantics. This project builds on a video analysis framework that leverages state-of-the-art methods for object detection and action recognition in a unified formalism encoded in terms of a mathematical and statistical approach known as pattern theory. The video analysis approach can (i) handle structural variability of complex events without requiring large training data while exploiting easily available ontological information, (ii) overcome classification errors of machine learning classifiers of actions and objects, (iii) accommodate scene clutter, i.e. extraneous objects that do not in the activity present in the scene, (iv) and manage sequences of elementary events, all without retraining. The formalism allows for the easy incorporation of temporal, spatial, and logical constraints. This team has demonstrated this system on standard datasets used to benchmark performance in computer vision for human activity recognition tasks.

This project, from the University of South Florida (USF), continues their activities to strengthen Innovation and entrepreneurship within the university and region through their I-Corps Site. Innovation Corps (I-Corps) Sites are NSF-funded entities established at universities whose purpose is to nurture and support multiple, local teams to transition their technology concepts into the marketplace. Sites provide infrastructure, advice, resources, networking opportunities, training and modest funding to enable groups to transition their work into the marketplace or into becoming I-Corps Team applicants. This is a Type II I-Corps Site - Type II projects reside at institutions that have already received a Type I award. I-Corps Sites also strengthen innovation locally and regionally and contribute to the National Innovation Network of mentors, researchers, entrepreneurs and investors.

This project integrates efforts from the USF Research and Innovation's Vice President for Research (VPR) Office with participants from engineering, business, and medicine, and support from their patent and licensing office. This I-Corps Site reaches across the campus and filled a crucial gap by providing hands-on entrepreneurial training to scientists and engineers about how to think about translating an idea to actual use by society. This Site at USF accomplished a great deal during their first three years and "changed the culture" of the institution - 115 teams completed the Site curriculum, 42 teams completed their boot-camp, and they sent 25 teams to the National I-Corps program.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Individuals with sensorimotor impairment must often rely on others to help them perform common activities of daily living. The goal of this project is to improve independence and quality of life by creating an adaptive human-robot collaborative system (a wheelchair mounted robotic arm) that learns from example to assist its user perform instrumental activities in a way that requires minimal user guidance. The project has a novel intention recognition framework that learns user goals despite imprecision of telemanipulation cues provided by the user during object interactions. The project also implements and tests a novel form of shared control authority that adaptively allocates workload between the human and robot to optimally leverage the physical capabilities and cognitive resources of the user. By doing so, this project will improve the independence of individuals with sensorimotor impairment and increase the autonomy of a wheelchair mounted assistive robot, thereby advancing NSF's mission by promoting the progress of science and advancing the national health and welfare. The project will involve an educational component that provides training to graduate students in conducting research. The project also develops a hands-on exhibit in collaboration with the Tampa Museum of Science and Industry, allowing visitors to explore the field of rehabilitation engineering.

This research investigates new control methodologies that promise improved autonomy in the control of a wheelchair mounted robotic arm operated by individuals with sensorimotor impairment. The project addresses key steps in the dexterous telemanipulation of objects: object detection, classification, and affordance modeling; user intention estimation; user / robot workload distribution; and algorithm training through teleoperation. Methods include computer vision, machine learning, probabilistic graphical modeling, and human subject experimentation to develop and test the collaborative human / robot system during performance of activities such as opening/closing doors with pull and knob style handles, and fetching objects in an unstructured populated environment. Effective application of the technology promises persons with physical disabilities opportunity to achieve a high level of independence, dignity, and quality of life.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

The lack of a strong public understanding of artificial intelligence (AI) technologies represents both a threat to national security and an opportunity for the development of new approaches to cybersecurity education and research. The project intends to develop and support a series of publicly accessible AI challenge competitions aimed at facilitating the public’s understanding of AI technologies and cybersecurity. All challenges will additionally serve as large-scale data collection platforms, assisting researchers in better understanding the trustworthiness and interpretability of AI systems. Project plans include creating lesson plans and challenges designed simultaneously to broaden understanding of cybersecurity issues in AI and provide an on-ramp for students interested in AI and cybersecurity across K-12 and higher education. Overall, the project team hopes to contribute to and scale ongoing efforts to develop and administer freely available online curricula that fosters public awareness of AI.

There is a clear and urgent need to connect research, education, and workforce development efforts at the intersection of AI and cybersecurity due to the highly interdisciplinary nature of AI and machine learning challenges, as well as the rapid development and adoption of AI technologies. This project will address these urgent challenges through the following research questions: First, how do individuals identify trustworthy AI systems and outputs? Second, how do AI/Cybersecurity concepts align with current educational standards? Last, how can public understanding and trust of AI be enhanced further? To answer these questions, the project will develop scaffolded challenges at the intersection of cybersecurity and AI entitled “Deeperfakes” and “P0150N”. The project team will work with the non-profit AI Education Project, working specifically to include modules on deep fakes and the role of AI in the cybersecurity workforce. The Deeperfakes challenges will ask participants to verify a series of AI-generated images and media, explaining their reasoning for either accepting or rejecting various media as real or fabricated. Similarly, the P0150N challenges will pit students against simple AI systems modeling those that detect denial of service attacks. This will allow students to experiment with mechanisms to poison the AI and successfully carry out an attack. The challenges will also serve as data collection tools and produce a large, open dataset for AI and cybersecurity research. This dataset will provide AI and cybersecurity researchers with information on how people read “faked” algorithmic media as real or fabricated.

This project is supported by a special initiative of the Secure and Trustworthy Cyberspace (SaTC) program to foster new, previously unexplored, collaborations between the fields of cybersecurity, artificial intelligence, and education. The SaTC program aligns with the Federal Cybersecurity Research and Development Strategic Plan and the National Privacy Research Strategy to protect and preserve the growing social and economic benefits of cyber systems while ensuring security and privacy.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.