BigInsight seminar on Practical Privacy-Preserving Distributed Statistical Computation
1 like115 views
The document discusses the Norwegian Primary Care Research Network's IT infrastructure known as the SNOW system, which facilitates privacy-preserving statistical computations using distributed anonymous medical data from general practitioners (GPs). It outlines the system's architecture, objectives, and benefits, including improved efficiency for research participation, secure data handling, and the ability to conduct automated epidemiological analyses. The document also highlights potential drawbacks, such as the lack of established standards among comparable systems and challenges in validating computed statistics.
![Snow architecture
- enables coordinated computations on distributed resources
- a “collaborative Edge computing” infrastructure [1]
S Coord
S
S
S
S
ClientCoord=Snow Coordination server
S= Snow Server in local health institution
Source: [1] Shi W, Cao J, Zhang Q, Li Y, Xu L. Edge Computing: Vision and Challenges. IEEE Internet Things J. oktober 2016;3(5):637–46.](https://image.slidesharecdn.com/620180305pcrnitinfrastrucutre-final-180430111139/85/BigInsight-seminar-on-Practical-Privacy-Preserving-Distributed-Statistical-Computation-7-320.jpg)
![Edge computing
“Edge computing refers to the enabling technologies allowing computation to be
performed at the edge of the network”[1].
Beneficial when data is:
• To sensitive (health data)
• To big (genetic data)
• To competitive (data will expose profile of owner)
• +++
Source: [1] Shi W, Cao J, Zhang Q, Li Y, Xu L. Edge Computing: Vision and Challenges. IEEE Internet Things J. oktober 2016;3(5):637–46.](https://image.slidesharecdn.com/620180305pcrnitinfrastrucutre-final-180430111139/85/BigInsight-seminar-on-Practical-Privacy-Preserving-Distributed-Statistical-Computation-8-320.jpg)
BigInsight seminar on Practical Privacy-Preserving Distributed Statistical Computation
- 1. The Norwegian Primary Care Research Network IT infrastructure: The Snow system Johan Gustav Bellika Professor Nasjonalt Senter for e-helseforskning Professor II Institutt for Klinisk Medisin, Helsevitenskapelig fakultet, UiT Seminar on Practical Privacy-Preserving Distributed Statistical Computations 2018.03.05 Dr. John Snow (1813 – 1858)
- 2. Source: WMA Declaration of Helsinki. URL: https://www.wma.net/policies-post/wma-declaration-of-helsinki-ethical-principles-for-medical-research-involving-human-subjects/ http://chrisricecooper.blogspot.no/2015/02/photoessay-on-year-of-ram-by-asian.html Declaration of Helsinki, Article 6 The primary purpose of medical research involving human subjects is to understand the causes, development and effects of diseases and improve preventive, diagnostic and therapeutic interventions (methods, procedures and treatments). Even the best proven interventions must be evaluated continually through research for their safety, effectiveness, efficiency, accessibility and quality.
- 3. Declaration of Helsinki, Article 9 It is the duty of physicians who are involved in medical research to protect the life, health, dignity, integrity, right to self-determination, privacy, and confidentiality of personal information of research subjects. Source: WMA Declaration of Helsinki. URL: https://www.wma.net/policies-post/wma-declaration-of-helsinki-ethical-principles-for-medical-research-involving-human-subjects/ Medical research should be privacy preserving!
- 4. Objectives for the research infrastructure • Make participation in research projects easier and more efficient for the GPs • Reuse health data in a safe and privacy preserving manner • Complete research projects according to scheduled time and resources consumption • Recruit 90-110 GP practices • Cover 7,5 % of the Norwegian population
- 6. What is Snow? • A distributed system • Enables collection and reuse of anonymous medical data • Builds and maintains a national online epidemiology-model • Use the epidemiology model to provide automated IT based health services • Enable privacy preserving distributed computations on EHR data • Directed at research, quality improvements, audit, disease surveillance,… Source:http://upload.wikimedia.org/wikipedia/commons/f/f6/Vibrio_cholerae.jpg
- 7. Snow architecture - enables coordinated computations on distributed resources - a “collaborative Edge computing” infrastructure [1] S Coord S S S S ClientCoord=Snow Coordination server S= Snow Server in local health institution Source: [1] Shi W, Cao J, Zhang Q, Li Y, Xu L. Edge Computing: Vision and Challenges. IEEE Internet Things J. oktober 2016;3(5):637–46.
- 8. Edge computing “Edge computing refers to the enabling technologies allowing computation to be performed at the edge of the network”[1]. Beneficial when data is: • To sensitive (health data) • To big (genetic data) • To competitive (data will expose profile of owner) • +++ Source: [1] Shi W, Cao J, Zhang Q, Li Y, Xu L. Edge Computing: Vision and Challenges. IEEE Internet Things J. oktober 2016;3(5):637–46.
- 9. The computing entities • The individual computing process – an “agent” • One instantiation at each participating Snow server • One unique communication address for each agent: agent-user@snow-server-domain/mission_id • Agents communicates among each other using XMPP messages • Coordinated computations: “Missions” of multiple agents: • One “main” coordinating agent • Multiple computation agents performing computations in parallel
- 10. Agent distribution scheme (Collaborative computations at the edges) Snow coordinator Main agent Snow server Snow server Snow server Health network Comput. agent Comput. agent Comput. agent Health institution Health institutionHealth institution
- 11. • A small computer that fits everywhere • Snow server software is pre-installed • Very easy installation • Remote system administration by the Snow team at UiT / NSE • Remove the risk of affecting the stability or performance of operation critical IT systems – the electronic health record system • All data in the box is pseudonymised, both patient and GPs • Agents compute on the box Snow appliance box: The nodes of the network 11
- 12. Data flow in PCRN Internet Secure health net GP office 1 Snow GP server EMR GP office 2 Snow GP server EMR GP office 3 Snow GP server EMR Aggregated data/statistics Snow coordinator server PCRN net portal • Distributed data analysis • Establish projects, invite GPs, initiate data extraction etc PCRN internal data • Epidemiological analyses • GP and patient data • Consultation statistics PCRN CN Safe haven for data Research data set (individual patient data)Secure data storage for research data set and advanced data analyses Local net = data storage
- 13. Using secure multiparty computations to support research in primary care
- 14. Virtualdataset Creating a virtual dataset with Emnet/Snow Researcher/PCRN staff Coordinator Def Def Def Def Clinical practice 1 Clinical practice 2 Clinical practice 3 Aimed at: 1. Make participation in research projects easier and more efficient for the GPs 2. Support researchers in inclusion of sufficient number of patients in clinical research 3. Support article 9 in Helsinki declaration: Privacy preservation
- 15. GP tool to identify the eligible patients
- 16. Virtualdataset Distributed statistical computations with Emnet/Snow Clinical practice 1 Clinical practice 2 Clinical practice 3 Researcher/PCRN staff Coordinator Query Query Query Query Result Secure multi-party computation (SMC) Aimed at: 1. Support researchers in inclusion of sufficient number of patients in clinical research
- 17. Report database Virtualdataset Automated processing Clinical practice 1 Clinical practice 2 Clinical practice 3 Coordinator Query Query Query Resultat Secure multi-party computation (SMC) PCRN interne data Aimed at: 1. Supporting article 6 in the Helsinki declaration: Continuous evaluation
- 18. Benefits • Centralised resources as PCRN staff/researchers can help GPs become more efficient in research. • Knowledge about the patient populations can be generated directly from the distributed sources, spanning administrative borders as municipalities, regions, countries and continents • Aggregated (non sensitive) statistics can be produced automatically directly from the sources.
- 19. Drawbacks • Two other comparable approaches exists, no standard established • How to validate correctness of computed statistics is an open research question