BC Platforms Pioneering TREs and Federated Data Analysis
BC Platforms Pioneering TREs and Federated Data Analysis
BC Platforms Pioneering TREs and Federated Data Analysis
Page 1
33 Million+ 17 Data
Patient lives in our catchment
459,614 Partners
area (BCRQUEST.com + Subjects with Genomic Data
EU, Americas, Asia-Pacific
Extended network)
and Africa
From Data To Health
Page 2
Page 3
No holes to firewall required, with All data kept behind data custodian firewall Firewall hole allowing access from public
data writing out blocked No access from outside to data server cloud, and allowing data writing to the public
cloud
Not possible to write data outside of data
server 3rd party has access to account and
encryption keys
Result of security breach/attack Attackers cannot move any data out from the Attacker can copy all data out (GDPR violation),
system by using for example a compromised research
account, man-in-middle attack, attack on vendor
system running in public cloud, or dishonest
researchers
Only workflow parameters shared — In TRUE federated analysis, only analysis Researcher able to run any workflows from
NOT workflows parameters are shared from the outside the outside
Result of security breach/attack All workflows used are approved by data Researcher can run workflow to copy all the
custodians. These workflows are proven to data (data-copy model)
produce GDPR non-personal classified results Executing workflows from outside permits
on any parameter values attacker or man-in-the-middle to execute
Possible attacker can therefore only access any workflow (e.g., data-copy, or delete-all
GDPR non-personal results workflow)
All inputs and sent results are logged for Can steal or destroy all the data from all data
auditing by data custodian — possible custodians
violations can be detected
Data server pulls and validates Data server pulls analysis parameters and Vendor server in public cloud pushes any
parameters performs comprehensive tests of parameters analysis parameters (and workflows) to data
value validities before executing analyses on server. Possible parameter value checks are
data server done on vendors’ server in public cloud
Result of security breach/attack Only validated data parameters are used for As parameter values are checked outside the
analyses data server, ‘man-in-the-middle’ attacker can
send malicious parameters (e.g., pieces of code),
causing workflows to malfunction or can copy
data out from the system
Manual check of all results/files In non-federated analyses, researcher stores Researcher can run any analyses or workflows,
final results in quarantine folder for manual and download any results files
check and acceptance
Result of security breach/attack Researcher can only download manually Researcher can download results that
verified results that do not violate patient compromise patient privacy, or are not in
privacy, and which only are in-line with scope of approved research plan, without any
approved research plan control
Note 1, as gathered from publicly available patent descriptions and similar documentation.
From Data To Health
Page 4
FinnGen
Initiated in 2017, the FinnGen project is a large public-
private partnership aiming to collect and analyse Finnish PRIVASA project — differential
privacy approach for analysing
genome and health data from 500,000 Finnish biobank
sensitive patient data and secure
participants, funded by the Finnish government and
federated AI model training
pharma companies. It is one of the very first personalised
medicine projects at this scale. BC Platforms delivered a
secure, audited BC|RQUEST system for performing data
availability queries based on clinical and genomic data,
available to all FinnGen researchers.
From Data To Health
Page 5