Improving Security: FortiVid and FortiDoc
Video surveillance systems, such as FortiCamera, help keep an eye on valuable assets, customers, employees and suspects by providing video feeds from every conceivable situation and location. The footage can be retrieved, as requested, to investigate or analyze a certain event. In other cases, an operator watches the live video and reacts immediately to unconventional activity.
For both cases, tools are available to facilitate the process. Motion detection can draw attention to events that might include suspicious behavior, or allow for a quicker browsing of hours of footage. Traditionally, facial recognition is largely done manually. In a similar manner to how a human guard would recognize a face from previous videos, an algorithm could carry out the same task, ultimately more accurately, faster and without losing concentration.
Dr. Mohamed Abdel-Mottaleb, professor and chair of the Department of Electrical and Computer Engineering at the College of Engineering, received a research grant from Fortinet to develop analytic tools that can remember faces of individuals and detect them again in other cameras and at other times.
Automated facial recognition (AFR) technology is vital for building automated surveillance systems. However, incorporating AFR into current surveillance systems poses a plethora of issues in unconstrained environments where there can be facial images with different viewing angles, varying lighting conditions, low resolution, and moving targets.
To target such issues, Abdel-Mottaleb explains that “the system will include components that identify frames which are more suitable for facial recognition from the video stream, use super-resolution methods, if possible, to obtain higher resolution images of the subjects’ faces, and use these images for learning and for identification.”
“The goal of this project started with a literature survey of the available methods for human identification using low-resolution images,” says Abdel-Mottaleb. “This was done in collaboration with Dr. Westendrop from Fortinet; we established the system requirements and a high-level design of the system.”
An initial prototype of the system is currently in development in order to show the benefit of this technology in automating the surveillance process. The emphasis will be on making the system more robust and accurate.
The second part of the research project consists of creating an algorithm, named FortiDoc, that will ultimately protect electronic medical records (EMRs) on a real-time basis. An EMR is a digital version of a paper chart that contains a patient’s medical history. An EMR is mostly used by providers for diagnosis and treatment.
The team that will execute this phase of the project consists of researchers from the Department of Electrical and Computer Engineering and the Miller School of Medicine.
Although there are already systems – based on overnight analysis – for protecting EMRs, this project focuses on real-time detection of privacy violations to then stop them in real-time.
“As with the other half of the project,” explains Abdel-Mottaleb, “a comprehensive literature survey will be undertaken to understand the pros and cons of available systems that are focused on securing EMRs. Then, we will develop a high-level design for the new system by establishing system requirements in collaboration with the medical team and experts from Fortinet.”
The project is officially titled, “Toward FortiVault, FortiDoc and FortiVid.”