We wanted our contextual smart home to be able to make sense of the current context and recent activity inside and around our smart home, so we have developed algorithms to determine a current 'threat level', which can then be used to adapt its behaviour and make it more intelligent.
Rather than invent a new scale for the threat level, we have adopted the one used by the UK government security services and modified it slightly:
The baseline threat level is set based on a number of factors, such as the current mode of our smart home, mains power being present, Internet availability, whether it is dark outside, the time of day, the current local weather, the forecast weather, current occupancy, doors open or unlocked, etc.
This baseline is then combined with an activity based assessment, to give a real-time view which is then part of the wider pool of context available to our smart home.
By definition, our contextual smart home "sees" everything that happens within it and around it. Each 'significant' event (as determined by our smart home) is captured and stored along with:
This will include things like the side gate opening/closing, shed door opening/closing, movement sensors in our garden and around the outside of our house and IP camera motion detection and face detection/recognition. It will also include things like the drive and porch being occupied, the letterbox sensor being activated or the door bell being pressed.
From this information a real-time threat level assessment can be made by looking through the series of stored events. This includes movement and occupancy and presence. It also includes all security and safety sensors.
Our smart home can hold a 'conversation' with family members and will also provide updates via notifications and voice announcements, when it thinks we need to be informed about threats or changes in threat level.
When the threat level hits level 5, the alarm internal alarm will be enabled.
The current threat level is something we can also query via our smart home's AI:
Our contextual smart home will change its behaviour based upon the current threat level. This may mean changes in lighting, the way IP security cameras operate, etc. It may even sound an external alarm automatically.
The actions taken are the main focus of our research in this space now, as threat level assessment seems to be working really well. Our @smartest_home really is a real-world test bed for this kind of research and it is only through trying things our in a real, lived-in home that we can identify new things to test and improve its operation.
Actions don't have to be local. Our smart home can reach out to known/trusted persons and services and this may include official organisations and the emergency services (if registered). In an assisted living environment, this could also include care service providers and friendly/helpful neighbours.
Our research in this space has been fascinating so far. It is making us think deeper about the smart home and what might be a threat to it. We are continually running experiments and new algorithms, to see what works well and what is less successful. This is likely to be a long-term piece of research.
Our contextual smart home occasional tweets the current threat level and daily high and low values (if different to current level).
