Minut, a Swedish startup that has developed a camera-less home security device that it claims protects privacy better than competitors, has raised $8 million in Series A funding.
The round is led by KPN Ventures, with participation from international energy and services company Centrica. Existing backers Karma Ventures, SOSV and Nordic Makers also followed on, bringing total funding for Minut to $10 million.
Founded in 2014 and headed up by CEO Nils Mattisson, who spent seven years in the Exploratory Design Group at Apple, Minut wants to make home security monitoring more affordable, but in a way that doesn’t compromise on privacy. To square that circle, so to speak, the startup’s IoT device is camera-less (in the traditional sense), and instead relies on other sensors, including infrared motion detection and a microphone. Crucially, the real-time data captured to determine if anything untoward is taking place in your home is processed on the device itself rather than being shared to the cloud.
“Feeling safe shouldn’t be a luxury, or come at the cost of privacy,” Mattisson says. “Until recently, the most affordable solution for home security and monitoring has been Wi-Fi-connected cameras, but people don’t want or trust them in their homes.” This realisation has seen privacy be the driver of Minut’s design decisions from “day one,” and is why the company was one of the first device makers to do machine learning “at the edge of the network.”
“This approach is technically much more challenging than recording sounds and sending them to a back-end for analysis like an Amazon Alexa or Google Assistant, but it enables us to identify events, such as a window-break or the presence of people, without ever actually recording any sound,” explains Mattisson. “Features are instead extracted from sensor data in real time and analysed on the device. When the local neural network recognises that something can be an event, only the extracted fingerprint then gets sent to a global classifier that can do a deeper and more accurate assessment. It’s not possible to reconstruct the sound from the fingerprint.”