Well, it’s official. The Internet of Things (IoT) has arrived. But without an "IoT privacy playbook" in our back pockets, questions continue to loom about the adequacy of existing privacy frameworks in this new paradigm of computing. Will the frameworks hold up? How can we preserve and manage privacy in a world of pervasive sensing and ubiquitous computing? In exploring these questions, one will inevitably come across a slew of commentary about the privacy-invading aspects of IoT applications, and what some critics have regarded as a “privacy nightmare.” Yet what is often overlooked in those discussions is the fact that one of the key technical tenants of the IoT infrastructure actually aligns and is not necessarily at odds with a key tenant of data protection.
Enter edge computing.
The key driver behind the IoT is edge computing, also known as fog computing. IoT networking giant Cisco Systems hailed “fog computing” as one of the six essential “pillars” of the IoT system, predicting that by 2018, 40 percent of all IoT-created data will be processed in the fog. So what’s the big deal with the fog, and how does it relate to privacy?
Fog or edge computing embeds computing power throughout a network instead of in a central cloud. Previous computing models rely on collecting in-field data and then transferring it to a central data center to run analytics. This won't be the case in the emerging IoT where data generated by "smart things" will be processed and stored on the "things" themselves. For example, instead of gathering data for analysis later on, smart traffic lights will analyze the data as it is being collected and make immediate decisions to improve the flow of traffic on a busy street.
As The Wall Street Journal’s Christoper Mims put it, “Whereas the cloud is 'up there' in the sky somewhere, distant and remote and deliberately abstracted, the 'fog' is close to the ground, right where things are getting done. It consists not of powerful servers but weaker and more dispersed computers. In appliances, factories, cars, streetlights and every other piece of our material culture."
Edge analytics is not a new concept; it’s just becoming more relevant as the number of connected devices expands. We benefit from edge computing on our smartphones. One of the reasons why mobile apps have become a predominant way to do things on the Internet is because some of the data and processing power is handled within your device. Shuttling every piece of data back to a data center requires a lot of bandwidth. As the number of sensors and “connected things” increases, edge computing will reduce the costs of data transmission, remove network delays and increase the speed of analytics. Now here’s where things get interesting from a privacy perspective. This is all simultaneously better from a privacy standpoint because edge computing minimizes the transmission of raw data such as MAC address, video feeds, still images, etc., to the cloud, and instead, applications will leverage "events," analytic triggers and metadata.
This is data minimization in practice.
Take, for example, video sensors. Video is one of the fastest growing types of data. With edge analytics, only the relevant and specific information extracted from a video feed, based on the detection of an analytic event, such as a car collision, is relayed to external systems—not the raw video itself. If a street camera captures video of a traffic accident, rather than transferring all of the video for analysis it can be processed on the edge and then generate an alert to the appropriate authorities that an incident has occurred along with information about the speed of the collision and the position of the wreckage.
Further, edge storage eliminates the need for bulk video archiving. Rather than recording and back-hauling video to an external storage drive, a trigger-based approach isolates recording to specific events of interest and stores the clip of the event directly on the edge for a finite retention period. Edge computing also might reduce challenges posed by the growing use of police dashboard and body cameras, which generate large volumes of audio and video recordings that have both privacy and storage implications. Edge computing cameras could analyze video feeds on the fly and only send relevant data when needed.
Not only does “privacy at the edge” align with one of the core technical tenants of the IoT, it also aligns with the “privacy on the fly” approach that Deirdre Mulligan and Kenneth Bamberger posited in 2011. This approach focuses on the governance of privacy through flexible principles and for continual iteration of data use and deletion policies.
It goes without saying that a new paradigm in computing is unfolding in front of our eyes. Could it be possible that emerging technologies underpinning the IoT architecture could avoid the "privacy nightmare" some anticipate?