Biometric recognition (for example, fingerprint recognition) offers many advantages over traditional methods (such as passwords or security cards) for identification and verification required for secure access. The algorithms enabling biometric recognition can be computationally intensive, but users of biometrics systems require identification and authentication processes to be rapid, convenient, and accurate. In this proposal, the team?s specific focus is on creating a fingerprint liveness algorithm (which detects if a fingerprint presented is live skin or a spoof) to improve accuracy and security of fingerprint detection. By enabling this algorithm to be calculated on the cloud, the team will be able to improve speed of execution and/or accuracy of scans and make the overall authentication system more feasible for use in different contexts.
In this proposal, the team proposes to undertake a project involving the transfer of parallelization of biometric applications on many-core platforms from research to industry. The benefits and uses of many-core computing are well studied already, as are advanced biometrics algorithms. The work being proposed will apply research related to many-core computing and parallelization to the field of biometrics in order to demonstrate performance improvements for a viable product. Specifically, the team will focus on a fingerprint liveness detection algorithm that can be used in conjunction with identification or verification. The work will help improve the security offered by such systems, and will demonstrate that established methods and concepts in many-core and cloud computing can be applied to this field.
This award was for a project addressing technology transfer from research to industry. Specifically, the project assessed the feasibility of commercializing parallel processing on many-core platforms for use in the biometrics industry. The potential impact of the project was adaptation of biometrics algorithms toward many-core computing platforms, with improved performance and efficiency. The goal of the project was not to find a way to get technology to market; it was to determine whether there was a commercially viable opportunity to do so. In order to accomplish this, the project required analysis of potential markets and critical assessment of the viability of commercializing the technology. The project involved researching the biometrics industry as well as possible alternatives. It also included interviewed over 100 customers to achieve both breadth and depth in determining customer receptiveness toward a commercial offering based on the use of many-core processing for biometrics applications. Finally, the project also involved synthesizing information obtained from this research to identify customer needs and to determine whether commercialization of many-core technology in biometrics would fill a specific need, which would indicate that there was a feasible opportunity for commercialization. Upon completion, the outcome of the study was the conclusion that the benefits of the technology were undeniable, but the current industry environment was not ripe for introduction of the many-core processing for biometrics as a product or service. This provided the valuable information that while the technology was in demand in multiple settings, demand was not consistent. Players in the biometrics industry are interested in the technology, but because it is so central to their business, many prefer to develop it in-house rather than to outsource it to a firm such as the one this project considered. As a result, the decision was made not to commercialize the technology at the time. While there is promise for its use in the industry in the future, and there are numerous further research opportunities, it was not the right time to transition it to industry in the form of a commercial venture. The benefits and uses of many-core computing have been well studied, as have advanced biometrics algorithms, although there is ample room for additional research. The biometrics algorithms studied are used for security purposes, such as identification or verification. The many-core computing techniques studied have been shown to improve performance and efficiency in many settings. Therefore, this project focused on applying many-core computing to biometrics algorithms. The work had the potential to improve the security offered by such systems by allowing for faster or more accurate operation. This would allow biometrics applications to take advantage of other advances in technology, and to perform better within a given set of constraints. While the results indicated that this particular application was not a feasible commercial opportunity at the current time, they nonetheless indicated that established methods and concepts in many-core and cloud computing can be applied very effectively in a variety of scenarios in the biometrics industry.
