TSD: a Secure and Scalable Service for Sensitive Data and eBiobanks Gard Thomassen, PhD Head of Research Support Services Group University Center for Information Technology (USIT) University of Oslo

Outline Sensitive Data TSD setup, solutions, demo, status and future How to get on board Some discussion topics Q&A

What is sensitive data? Norway : Personal Data Act §2, point 8 –race/ethnic data, political opinion, philosophical and religious beliefs, the fact that a person has been suspected of, charged with, indicted for or convicted a criminal act, health, sex life and trade-union membership

Who has sensitive data Almost everyone

TSD launch in Computerworld 16/5-14

Norsk KreftGenom Konsortium Sammenliknet med den hardware vi benyttet fram til overgangen til TSD, som vel kan karakteriseres som en middels brukbar tjenermaskin, med 64 kjerner, kan vi med TSD oppnå en teoretisk hastighetsforbedring på 30X. I tillegg til dette kommer at vi har opitmalisert vår analysepipeline, ved at vi har parallellisert flere trinn. Tidligere ville en sekvenseringsanalyse på 48 svulst/normal-par resultert i kjøringstid på to-tre måneder minimun. Vi kjørte nå denne uka på TSD det samme på to dager og noen timer. Altså forsiktig sagt en dramatisk forbedring. Prof Eivind Hovig, NCGC

Teknisk ukeblad & e24, 5/5-14

Uniforum

TSD Pilot

System requirements Security, isolation and access control as given by law Large storage capacity Multi tenant (multiple users) High performance computing (HPC) resource High bandwidth Easy to maintain and operate Easy to use and “practical” (also for audio and video) Some freedom within a confined user space Accessible from anywhere through proper mechanisms A variety of software and public data-sources must be available Windows and Linux support (server/host-side) Data collection services Data sharing services

Setup, solutions and status

System outline Gateway HPC - ColossusVM-server Storage Internet Secure encrypted network to special high volume data production sites 1 (project) 1 (storage area) n 1

TSD demo Check out the login help pages, all projects can use PCoIP and ssh+RDP to access windows from 15/8-15 Thinlinc (html5 based ) for access to linux machines in TSD will be enabled by August Help pages :

Data import and export using TSD File lock server Virtual file lock server Virtual project- server File lock HD Project HD TSD NFS mount 2 Data copied here by sftp (2-factor authentication) encrypted data if sensitive 1 4 3

Data collection using TSD “Nettskjema-minID” “Nettskjema-minID” Nettskjema homepage minID Project VM Project disk File lock Encrypted XML (PGP) TSD

Security details OATH TOTP 2-factor authentication –Smart phones or programmable hardware tokens –2-factor absolutely everywhere Import/export is under strict control No open connection to the internet All administration happens from the inside Strong separation between projects Hardened FreeBSD gateway and firewall Encrypted backup, one key per project Sys-admins are single users (traceability) Sys-admins have to use same authentication process

Homepage Risk evaluation etc : mer-om/systembeskrivelse/

Projects g/sensitiv/mer-om/kunder/

TSD status > 90 research projects > 350 users Secure storage (> 1 PiB on disk) Secure data analysis Linux or windows hosts (> 250 VMs) Secure import and export Web-based data harvesting HPC cluster (>1500 cores) Postgres DBs Video and sound display

Capabilities enabled by TSD Large scale NGS research on human genomes Large scale medical imaging studies Large scale studies with web-based data collection Cross border data-collection Off-site analysis of sensitive data Secure storage for verification of published research Electronic consent, soon Free storage from Norstore, soon free CPU?

Nordic collaboration opportunities Laws are fairly similar (Norway very strict) Difficult to exchange sensitive data for research One should learn from each other as these systems demands very special IT-knowledge Services development and system-administration know-how is non-sensitive and may be shared Building TSD addressed many novel security questions in a University setting to be learnt from Large DBs/registeries of health data may enable very interesting research in the future TSD is involved in the NeIC-based Tryggve project We are happy to collaborate!

Main collaborators on TSD Collaborators Norwegian Storage Infrastructure (NorStore) Norwegian Genetics Analysis Platform (GenAp) Norwegian Dietary Registry (Medical Faculty) Institute of Psychology (Faculty of Social Sciences) Norwegian Cancer Sequencing Consortium (NCGC) Reference group Oslo University Hospital, NorStore, Regional Ethical Committee, National Institute of Public Health, Norwegian Cancer Registry, Research Network at OUS, Elixir Norway, NCGC, GenAP, Institute of Psychology.

Future of TSD - main topics More on how to handle video and sound –harvesting –management –metadata –analysis Journal system for Psychologists (Univ of Umeå collaboration) Biobanks computing resource ? Thinlinc (August 2015) and PCoIP (15/8-15) VMware and VDI infrastructure for video Galaxy inside TSD Elixir helpdesk connected to TSD Hosting docker containers Invariant storage of research data (connected with Cristin ?) National eInfrastructure investment in TSD ??

Use of direct identification inside TSD Disclaimer / instruction : “Prosjektleder er ansvarlig for at man benytter direkte identifiserbare data så lite som overhodet mulig inne i TSD. Ved bruk av Nettskjema + minID/BankID skal personnummer vaskes bort og koplingsnøkkel lagres slik at den er tilgjengelig for så få prosjektdeltakere som mulig“ hjelp/secure-nettskjema/index.html

Unencrypted large datasets -> TSD Controlled end-end networks using ACL´s SFTP with 2-factor No open internet connections in any endpoint

PCoIP (encrypted and 2-factor) TSD then needs port 22 for sftp/ssh and port 4172 for TCP and UDP Directional control of c&p

Thinlinc Linux Remote Desktop protocol over html5 (https) with 2-factor authentication BSD firewall in front Same in principle as PCoIP setup : –“Login server” –“Connection broker” –“Virtual project machines / servers” –Can have directional control of c&p

Thanks to Project group / developers IT-dir Lars Oftedal Hans A. Eide Märtha Felton Reference group Administration / associated