Page 170 - ITU KALEIDOSCOPE, ATLANTA 2019
P. 170

2019 ITU Kaleidoscope Academic Conference




            4.  PRELIMINARY ARCHITECTURE PROPOSAL             the  functionality  (e.g.  access  control,  auditing,  context
                                                              integrity checks) to use the data to be shared with the health
           The architecture proposed in this paper is framed for a very   service providers; this is achieved through the cloning and
           specific  digital  health  scenario,  specific  requirements  and   migration of the DataAgent from the PHIX main container
           specific  context.  The  scenario  is  that  of  sharing  personal   to the service provider containers, and provides inter-agent
           health information data (e.g. health indicators collected on   communication and synchronization between the associated
           personal  monitoring  devices,  historical  health  records  and   DataAgents. The data owner has control of his community
           digital phenotypes [34]) with a health service provider, and   of DataAgents with the ability to gain visibility of where his
           ensuring  legibility,  agency  and  negotiability  in  the   data has been shared, to understand the specific utilization of
           interaction between the individual and their data. A subset of   their data and to control the use of specific DataAgents, for
           the requirements detailed in the previous section can be met   example, updating permissions and access control, revoking
           and  implemented  with  standard  techniques  and  solutions.   and killing shared agents.
           For example, some of the requirements around data privacy
           can be handled using information security techniques, such
           as public cryptography systems [35], as has been the practice
           for  say  HIPAA  compliance  and,  more  recently,  GDPR
           compliance.  However,  there  are  specific  requirements
           associated  with  the  HDI  imperatives,  specifically
           negotiability,  that  give  motivation  for  the  architecture
           proposed  in  this  paper.  In  particular,  the  architecture
           addresses  the  requirement  for  enhancing  the  control  that
           owners of data retain over their data once the data is shared,
           and  ensuring  that  the  dynamic  contextual  constraints  are
           enforced on the subsequent use of the data.

           The  proposed  architecture,  Personal  Health  Information
           eXchange  (PHIX),  is  based  on  the  multi-agent  systems
           (MAS)  paradigm.  As  a  candidate  implementation  of  this
           architecture,  the  JADE  multi-agent  system  platform  is
           considered  [36].  JADE  is  a  framework  to  develop  agent
           applications  in  compliance  with  the  Foundation  for
           Intelligent  Physical  Agents  (FIPA)  specification  for
           interoperable multi-agent systems. FIPA aims at providing a
           reference  model  for  the  implementation  of  highly
           interoperable  complex  agent  systems.  The  specification
           defines a minimum set of key agents that are necessary for   Figure 3 - Health informatics architecture proposal
           the  operation  of  MAS  platforms;  these  include  the  Agent
           Management  System  (AMS)  –  which  provides  for  the   Within this architecture, as per the defined scenario, service
           management and control of other agents on the platform; the   discovery is primarily handled via the use of the DF through
           Agent Communication Channel (ACC) – which provides the   which  the  health  service  providers  publish  their  details.
           communication mechanism between agents on the platform;   Individuals who need to share their data (i.e. via cloning and
           as  well  the  Directory  Facilitator  (DF)  which  provides  a   migration) would similarly employ the DF to query the data
           yellow  page  services  for  the  agent  platform.  Besides   for  the  relevant  health  providers.  The  key  feature  of  this
           providing the basic features that are specified in the FIPA   architecture is to bundle personal health data along with the
           specification,  JADE  provides  a  JAVA-based  distributed   functionality to manage its use in various contexts into the
           agent  platform,  with  transport  mechanisms  for  inter-agent   DataAgent, which is under the control of the data owner. By
           communication,  automatic  registration  of  agents  with  the   exploiting the agent mobility and migration feature of MAS,
           AMS, a GUI for the management of the agent platform, a   sharing of data is not associated with relinquishing control
           library of FIPA interaction protocols, as well functionality   over  the  data,  the  DataAgent  enforces  the  contextual
           for monitoring the interactions between the agents [36].    constraints, as well as the dynamic access controls decided
                                                              by the data owner.
           PHIX consists of the core MAS platform, distributed agent
           containers hosted by the  health service providers, and the        5.  DISCUSSION
           DataAgent, which encapsulates personal health data, as well
           as the functionality associated with the use of the data. The   While  the  PHIX  architecture  fundamentally  explores  and
           key  element  of  this  architecture  is  the  DataAgent  which   illustrates  the  operationalization  of  ethical  principles  and
           functions  as  a  mobile  virtual  data  double  that  allows  for   values, such as the HDI imperatives, in technology solutions,
           individuals’ health data to be secure packaged, using relevant   it also proposes a technical solution for the specific challenge
           knowledge  representation  standards  and  ontologies  (e.g.   associated with enhancing access control to and contextual
           triple-based RDF or RDFS); allows for the data along with   integrity of personal health data. Some of the solutions that




                                                          – 150 –
   165   166   167   168   169   170   171   172   173   174   175