Privacy-protecting tech that keeps the security benefits of facial recognition
Advances in facial recognition (FR) make it one of the fastest growing industries. The technology alerts security services to the presence of known criminals or terrorists, improves authentication for access to buildings or accounts, and can help track the habits of shoppers/visitors for the targeting of real-time, location-specific engagement options. Equally, FR can be used to monitor individuals and is also of interest to malicious users for the collection of data for hacking or identity theft. In short, as well as increasing security, FR can also compromise public safety, while invading personal privacy to an unprecedented degree. With the help of EU funding for SMEs, in October 2018 D-ID launched their disruptive solution for privacy-protecting facial images, that preserves their security utility. Several large-scale implementations of the technology have yielded promising results in terms of efficiency (i.e. protection against facial recognition engines) and customer satisfaction.
Keeping FR in check
“We recognised that with cameras now everywhere, if image sharing is combined with emerging FR technology, our most sensitive biometric asset – our face – may be compromised without our knowledge or consent. We wanted to build a solution to keep FR in check and ensure that people retain control over their identity and privacy,” says Sella Blondheim, co-founder and COO at D-ID. D-ID uses advanced image processing and deep learning techniques to recreate any photo so that the face (or faces) in them, while perfectly recognisable to the human eye, look different to FR algorithms. The team tested the technology with some of the world’s leading FR engines, including Google, Amazon, Microsoft and Facebook. They presented the FR engines with two sets of images. The first, an original, unprotected version of one, or several, reference images. The second set included a protected (de-identified) image representing the same reference image/s as in the first set. In all the tests, the FR engines were able to match the original image with its references, but they were unable to do so with the protected image.
Privacy by Design
The EU’s GDPR privacy regulations cover many aspects of data security, including the collection, handling and storage of biometrics and personally identifiable information. It is the responsibility of individual organisations to apply the guidelines. With fast-growing FR, companies are increasingly at risk of falling foul of the regulations and so facing not only heavy fines but also lawsuits, not to mention reputational damage. “D-ID is the only company in the world that challenges facial recognition this way. We do not accept the notion that we have to sacrifice our privacy for safety and security,” says Blondheim. D-ID can be used for a number of Privacy by Design initiatives such as protecting photo databases of customers and employees or protecting privacy in digital onboarding and know-your-customer processes. The team continue to enhance D-ID’s algorithm, making it even more accurate, and are miniaturising the system for handheld devices such as cameras and phones. They also recently launched a new solution for the anonymisation of video and still images which could open up the market for visual content, which is growing fast (with initiatives like the Internet of Things) but was previously restricted by privacy concerns.
Keywords
D-ID, facial recognition, privacy, security, GDPR, surveillance, hacking, spy, algorithm, image processing, photo
