Computer Vision News - February 2018

Computer Vision News 19 UnifyID different sensors. Vinay and his team fully understand this apprehension. They passionately advocate this new philosophy of authentication while remaining extremely careful and honest about which aspects of sensor data is available to them. Using an amalgamation of ideas from cryptography and deep learning, UnifyID’s authentication pipeline operates in a differential privacy regime that obviates the possibility of them turning into a ‘ biometric honeypot ’. Vinay and his team want to make this the industry standard, knowing the importance of security regarding biometrics. They also insist that this approach is markedly different from the active biometrics field where theft of a person’s biometric such as iris or thumbprint might render that victim exposed for the rest of his or her life. Vinay explains that the development of their application has two phases; the product side and the company side. On the product side, they have their SDK ready for integration with other apps, and many of their partner companies already have UnifyID embedded. Their system also provides GPS services. They have developed their application using a two-pronged strategy . One is in terms of providing the confidence scores emanating from specific passive factors such as Gait, Bluetooth, etc. exposed via an API ensuring that users can handle the end product however they want. Each user may have their own security partners already with their own biometrics. They want to understand the level of confidence that they can get via gait or gaze, for example. The other is a more holistic solution which they offer as an SDK for many of the activity apps. They will provide authentication as an additional service. Currently, they are in the final stage of collaborating with a large smartphone device manufacturing company as well as a big banking corporation providing a solution for ATM machines. As far as the company is concerned, UnifyID just finished raising their series A from a top-tier venture capitalist firm in Silicon Valley. In terms of finance, they have made it through the feverish series A phase. They are good to go for the next few years in terms of money. On the algorithmic side, Vinay details the four main components of their machine learning pipeline. The first component , termed broadly as the signal processing component, entails Application UnifyID demo video from TechCrunch Disrupt Attacker rejection