Appliance Inference

Up until now this site has discussed the privacy implications of inappropriate access to data with a 30-minute granularity. Some companies are working on capturing data with a much finer resolution – as often as once per second. For example, look at the work of Onzo in the right-hand pane.

If energy consumption data is captured with a sufficiently fine-grained resolution then it is possible to match the pattern of energy consumption against templates corresponding to known electrical appliances that might be found in the home, in a process known as “appliance inference”. This could in principle allow:

  • Electricity bills itemised by appliance.
  • Detection of faulty electrical appliances, perhaps before they fail.
  • The temperature at which a washing machine is operated.
  • The time at which a dish washer runs.
  • Identification of old, inefficient appliances that could be replaced.

With advice derived from such information, a customer could make significant savings in electricity bills.

Of course, in the wrong hands this level of information could represent a gross invasion of privacy. Ideally the data should stay within the house.

At some point in the future it might be possible to create an “appliance inference applet”, which performs its appliance inference work within the smart meter, and reports the useful information to the customer without the need to export the raw data.

Or a variant of the “data export applet” mentioned in the previous tab could move compressed data securely and with the informed consent of the customer to a web site for further processing.

Appliance Inference for Assisted Living

We would like to finish with a final, elegant, suggestion that ties together the two threads of the Project Hydra work – assisted living and secure microcontrollers.

It is recognised that it can be useful to monitor the activity of some assisted living patients as the go about their daily activities. Have they got out of bed? Are they cooking and making themselves cups of tea? Are their bathroom patterns normal? Some of this monitoring requires dedicated sensors – bed occupancy sensors, for example. But it appears that appliance inference will be able to provide very useful input, at zero extra cost.

Therefore we propose the “assisted living appliance inference applet”.