IoT006 - Organicity Technical Deployment Project Brief

IoT006 – OrganiCity Technical Deployment

Project Brief v3.0

Document Sign-off  

Background

This project is initially funded by OrganiCity, an EU-funded project that offers a service for IoT experimentation, focussing on how “citizens, businesses and city authorities can work together to create digital solutions to urban challenges”. OrganiCity experiments take place in cities across Europe, and the Edinburgh CitySounds is one of 17 experiments that were approved in the second and final funding round coordinated by OrganiCity.   It is the intention that the UoE will continue to run the infrastructure and capture data after the OrganiCity funding period ends in March 2018, such that we have a complete 12 months of data.

The project will deploy six so-called Audio Capture Devices in the Meadows in Edinburgh to capture audio as a valuable first step in exploring and celebrating the richness of sounds in the city, benefiting from recent innovations in digital technology and network infrastructure. It will focus on how sounds from biotic (e.g., animals), abiotic (e.g. weather) and anthropogenic (human activity) sources, captured in a central urban greenspace, can inform community groups and citizens about biodiversity and health and well-being, as well as provide a unique resource for artists and data scientists. Wide-frequency audio will be captured in near real-time for secure, privacy-preserving processing and analysis. The Audio Capture Devices will be designed and constructed by the team so as to capture both ultrasonic and audible sounds and transmit them via WiFi to a central server for further processing. Given the nature of the deployment (microphones in a public space), the project will ensure that appropriate steps are taken to ensure adherence to data protection legislation, and this will include a ‘scrambling’ process to obfuscate any human speech captured by the devices, with the result that the content is unintelligible and no individuals could thereby be identified. Representative samples of this anonymised audio data will be made available to other registered OrganiCity experimenters. This project is led by Edinburgh Living Lab, a city-wide collaboration founded by the City of Edinburgh Council and the University of Edinburgh.  

Scope

This project will primarily focus on the technical deployment of the devices, the associated software infrastructure and the pre-requisite data protection, ethical and City of Edinburgh Council Park Services approvals needed prior to installing the Audio Capture Devices in the Meadows.  

Out of Scope

Engagement with researchers and external stakeholder groups (specifically Friends of the Meadows and Bruntsfield Links, Scottish Wildlife Trust and New Media Scotland) will not be managed by this ITI project.

Management of delivery of related CitySounds web site or sound installation using sounds captured from the Meadows.

Objectives and Deliverables and Success Criteria

Summarise - What we are aiming to do and deliver, by when.

What are we aiming to ultimately deliver that will be key to the measure of the success of this project - but potentially any significant other things along the way (that may outlive the project).

What are our objectives and what deliverables are part of them.

Driven by the Deliverables and Objectives. Do we have objective measures that we can point to to say the project has been a success (typically we said we would deliver X and we have).

Unless otherwise indicated, the deliverables are a success when accepted by the Project Sponsor and the Service Owner.

Requirements

Requirements are aligned with objectives and deliverables, and also the opportunity to realise the benefits.

Benefits

Partnerships

Despite its modest scope, the project has fostered the first ever research collaboration between a unique set of cross-disciplinary, cross-sector partners. Within the University, the collaboration spans from Information Services and Informatics through to Biological Sciences and The Reid School of Music.  Outwith the University, it comprises all the partners of The Edinburgh Living Landscape,[1] as led by Scottish Wildlife Trust; community groups led by and associated with Friends of the Meadows and Bruntsfield Links; and the cultural organisation New Media Scotland.

Expected CRD Outcomes

Talent

• Two undergraduate students have been co-opted into the project team to help construct the Audio Capture Devices and are broadening their skills base and understanding of the value of wide spectrum audio data.
• An Informatics PhD student has been responsible for the design and implementation of the voice scrambling technique, and has consequently acquired greater understanding of the state of the art in speech obfuscation as well as experience of participating in a highly interdisciplinary, cross-sector research project.
• A PhD student from School of Biology with a background in biodiversity is acting as an intermediary between the project and one of the stakeholder groups and is gaining increased awareness of the value of IoT and audio data in monitoring biodiversity.
• Via two community engagement workshops organised by the project, citizen science and voluntary groups are gaining tools for increased data literacy and awareness of the value of IoT and audio data in monitoring biodiversity.
• Our existing approach for audio capture is informing how we could build it into a suite of IoT tools and sensor data for incorporation into the science and maths curriculum for schools in the city region.  We are also exploring how selected audio could be made available for integration into the map resources being developed by the Edina DigiScience initiative.

Research

• The project team is preparing a paper on the project for the 3^rd International Conference on Smart Data and Smart Cities (Delft, October 2018).
• Members of the team, including external partners, are preparing a bid to the Nesta “Rethinking Parks” call.
• A PhD studentship in the School of Biology will be made available to carry out research on ultrasonic data being captured by the project.
• Tim Van Dam, founder of InternetOfLife (which uses IoT for ‘smart parks; and protection of endangered animals) is keen to deploy the CitySounds technology into a national park in Africa.  This would not only be an interesting and much rarer audio dataset captured for research but would also to help engage school children in IoT and biodiversity monitoring through audio, extending from their local context to something very exotic.

Adoption

Experience gained in the project will help inform and develop the design of our broader sensor network we are seeking to deploy as part of the City Region Deal.  A microphone with embedded intelligence on board[2] could be included within the composite sensor device that will be deployed at scale across the region.

The CEC Biodiversity team are interested in the potential for the proposed approach to provide an always-on, low energy IoT solution for 24/7 biodiversity monitoring.  This could form part of a suite of technology solutions to replace existing infrequent, expensive manual site-surveys.

We are in contact with a number of companies and organisations that are interested in exploring possibilities for:

• audio scene analysis for mobile devices, smart audio speakers, speech-enabled bots;
• detection of drones and gunshots
• sensors for noise pollution in smart city contexts; and,
• biodiversity consultancy and analysis of ecosystem services.

Datasets

The project will produce a unique collection of rich data, containing broad spectrum audio (0Hz to 96KHz) in FLAC format. Over the course of 12 months, we expect this to amount to 6Tb in total. This project-generated dataset has the potential to kick-start a cluster of new research activities within the University, including the development of statistical classifiers for the emerging field of “machine hearing”.

The availability of better audio classifiers will underpin efforts to embed more intelligence into microphone-enabled edge devices. This would greatly broaden the scope for deploying audio capture devices and allow them to be connected to a LPWAN rather than requiring Wi-Fi.  It is anticipated that such devices will be more manageable from a privacy-preserving perspective, and could thus be made more easily available to citizen science groups, as well as schools for a STEM curriculum topic centred around the environment and biodiversity monitoring.

As well as supporting the development of better tools for monitoring biodiversity, the dataset will allow us to build detailed maps of noise pollution and to recognise a range of different activities, including:

• sports events,
• fairground, festival and musical events,
• the presence of emergency vehicles,
• the occurrence of vehicle collisions. 

These have the potential to be incorporated into a digital map of the area of study.  Providing these as a live feed would be one area of further potential research/development.

On the cultural side, the data will offer a novel resource for visualisation by digital artists and creatives.

Entrepreneurship

The project was set up to support citizen science, and had no specific objective to stimulate entrepreneurship.  However, the approach taken and expected dataset provides a resource that can be made available for entrepreneurs and start-up/scale-up businesses to work with.

[1] https://edinburghlivinglandscape.org.uk

[2] As discussed further in the section on Datasets

Governance

Project will have these governance roles by default. Delete/Add/Change as appropriate.

Programme Governance

Project Board

Stakeholders

Resources Skills and Cost

Budget  

OrganiCity Funding for:

• Staff: 15.1k
• Infrastructure and other direct costs: 14.8k

ITI funding for this financial year to continue hosting the infrastructure and collecting data for the remainder of this financial year: 3.5k

Note: Circa 2k will need to go into the 2018-19 budget to continue hosting the OrganiCity infrastructure from August 2018 to March 2019 (to provide one full year of data).

All other effort beyond the OrganiCity funded period is core funded.

Priority and Funding  

This is a high priority project within the IoT Programme.

Funding is in place as noted under “Budget” above.  

Project Team

Quality of Project and Deliverables / Key Project Milestones  

Assumptions

1. It is assumed that the speech obfuscation process will be successful and independently verified.
2. It is assumed that the Privacy Impact Assessment will be approved.
3. It is assumed that the convenor of the Informatics Ethics Panel will verify that the proposal adheres to the School’s Ethics Procedure
4. It is assumed that City of Edinburgh Council will approve the siting of audio capture devices on specific trees in the Meadows.

Constraints

1. Adherence to Data Protection Legislation.
2. Audio capture devices can only be installed in locations confirmed by the City of Edinburgh Council

Risks

1. Reputational risk for the service:

Misunderstanding by the public of what the ACDs are actually doing, i.e. that they perceive they are under microphone surveillance and that we may be recording any conversations they may be having whilst in the Meadows, and potentially sharing them with third parties.  

2. Exposure of personal information when in service: 

Weaknesses in the technical implementation of the system, and potential for human error where unscrambled audio data which happens to contain discernible voice traces is inappropriately shared outside the boundaries of the secure Data Collector  

3. Theft (or destruction) of an ACD from the tree to which it has been deployed in the Meadows.  
4. Compromise of the Data Collector or other systems within the University network from a stolen/hacked ACD when in service.  
5. An ACD box could fall out of a tree and hurt someone after deployment.  
6. Someone may attempt to manipulate an ACD in the Meadows to record known speech and then attempt to show the voice scrambling scheme we are using is not effective, by producing a matching recorded "unscrambled" voice sample.  
7. Brute force password attack to user accounts on back-end servers.  
8. There is a risk that the PIA may not be approved which would prevent the project deployment going ahead.   Approval may not be provided by the DPO, the G&E AG, or the project\programme sponsor for some reason.  
9. There is a risk that Informatics Ethical approval may not be given, which would prevent the project deployment going ahead.  
10. There is a risk that resource availability will not be sufficient to meet the required timescales for the project.  
11. There is a risk that someone could be hurt while installing an ACD in a tree, replacing batteries in an ACD already in a tree, or during the eventual removal of the ACDs.

Issues

1. Cost for wireless receiver install was significantly more than expected.  
2. Wireless receiver coverage not as wide as expected  

Risk and Issues will be maintained in the Project Risks and Issues logs.

Previous Lessons Learned

Some experience has been gained with Privacy Impact assessments in previous IoT projects.

Dependencies

This project is not dependent on any other project

Communication

1. Project and Programme sponsor is part of the project team.   Project team meetings will be held as needed.
2. Monthly project updates reports will be published on the UoE IS Projects Website.
3. Internal email announcement to stakeholders on go-live.
4. Press release on go-live.
5. Notices place in notice boards in the Meadows.
6. Social media announcement from the UoE IoT Twitter account
7. Although not managed directly by the project, the following groups will be engaged regarding the project:
   1. The Edinburgh Living Landscape.
   2. Scottish Wildlife Trust.
   3. Community groups led by and associated with Friends of the Meadows and Bruntsfield Links
   4. New Media Scotland
8. Share the PIA with City of Edinburgh DPO.

For Run / Grow / Transform, Alignment with Strategic Vision, Digital Transformation, and Themes, refer to the IoT Programme Brief Addendum - Strategic alignment and themes doc

Service Excellence - Information and Security

1. Standard builds are being used for the servers with some additional CIS benchmark hardening applied. 
2. A review of the system will be carried out with the UoE InfoSec team.
3. Only UoE staff will be installing and supporting the servers and software.
4. A privacy impact assessment will be carried out and approved.

Service Excellence - Process Improvement, efficiency, quality and best practice (Social Responsibility and Sustainability)

Not applicable.