Embracing Opportunities of Livestock Big Data Integration with - - PowerPoint PPT Presentation

Embracing Opportunities of Livestock Big Data Integration with Privacy Constraints

Franz Papst 12 Olga Saukh 12 Kay Römer 1 Florian Grandl 4 Igor Jakovljevic 5 Franz Steininger 3 Martin Mayerhofer 3 Jürgen Duda 4 Christa Egger-Danner 3

1TU Graz Institute of Technical Informatics, Austria 2Complexity Science Hub Vienna,

Austria 3ZuchtData EDV-Dienstleistungen GmbH, Austria 4LKV Bayern, Germany

5smaXtec, Austria Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 1 / 13

D4Dairy

• Interdisciplinary research project for digitalisation in dairying
• Provide digital support to dairy management
• The four Ds stand for

▶ Digitalisation ▶ Data Integration ▶ Detection ▶ Decision Support

• 44 project partners

▶ 31 industry partners ▶ 13 scientifjc partners

• https://d4dairy.com/en/

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 2 / 13

D4Dairy

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 3 / 13

Introduction

• IoT has a big impact on agriculture
• Cheap sensors enable a paradigm shift in farming

▶ monitoring fjelds ▶ monitoring livestock ▶ …

• Potential use-cases for IoT in

dairying

▶ activity ▶ body temperature ▶ feed intake ▶ milk yield Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 4 / 13

No Legal Basis

• Farm data is in general not afgected by privacy regulations

▶ it should be more viewed as farmers’ trade secrets

• Same applies for sensor companies
• Communities push standardisation

▶ e.g., ICAR

http://www.molevalleyfarmers.com/mvf/info/general/smaxtec-animal-care-system

Valuable cattle data remains in silos.

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 5 / 13

Difgerent Perspectives

• Farmers

▶ have difgerent sensors on their farms ▶ which come with difgerent applications ▶ want a unifjed view of what’s happening on the farm

• Sensor Companies

▶ want to improve the utility of their products ▶ hesitant to share their proprietary data

• Veterinarians
• Regulatory Agencies
• Federal Agencies
• Cattle Breed Associations

Analysing data is key, but it has to be done with privacy constraints.

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 6 / 13

Our Contributions

• Survey of existing systems
• Propose a privacy-preserving data integration architecture
• Give examples of potential use cases

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 7 / 13

Existing Systems Overview

System Live since Architecture Data Storage Privacy Analytics NCDX 2015 centralised ✗ ✓ ✗ OFIS 2015 centralised ✓ ✗ ✗ JoinData 2018 centralised ✓ ✓ ✗ 365FarmData 2013 centralised ✓ ✓ ✗ Barto 2018 centralised ✓ ✓ ✗ ADA 2018 decentralised ✗ ✓ ✗ ODiL not yet decentralised ✗ ✓ ✗ HARA 2019 decentralised ✓ ✓ ✗

Existing systems exchange data, but do not use it further.

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 8 / 13

Our Architecture

Data Exchange

F A R M S Sensor Company

Federal Agency

Local Scope Global Scope

Data integration on a global scale is required.

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 9 / 13

Example

• Investigate relation between feed intake and milk yield

▶ from difgerent farms ▶ with difgerent equipment ▶ without disclosing those values

https://www.gea.com/de/products/ gea-free-stall-feeder-wic.jsp https://www.lely.com/gb/solutions/milking/taurus/

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 10 / 13

Privacy-preserving Data Analytics

• Secure multi-party computing
• Federated learning
• Data anonymisation and obfuscation
• Synthetic data generation

https://aircloak.com/data-anonymization-use-cases/

All these methods allow data analysis, while not exposing private data.

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 11 / 13

Applications and Benefjts for the Farmers

• Predicting missing and delayed data
• Anomaly and event detection
• Monitoring cattle health and welfare
• Autocomplete and verifjcation of manual input

https://www.allflex.global/livestock-monitoring-and-intelligence

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 12 / 13

Conclusion

• IoT has big potential for the dairying industry
• Privacy concerns are currently hindering potential use-cases
• We propose an architecture, which is able to analyse data in a

privacy-preserving manner

• This system enables better utilisation of data

Franz Papst (TU Graz / CSH Vienna) Livestock Big Data & Privacy 13 / 13