BLockchain Enabled Deep learning Data analysis

The technology that is being developed at the FIT within the BLENDED project
helps the European Space Agency (ESA) process images of Earth. The project
connects scientists across Europe in an effort to create a revolutionary platform
for distributed and, most importantly, secure data processing using artificial
intelligence that processes and analyses space data.

How can we share
results from space-based observation and maintain the integrity and security of
such data? In the following year, researchers from the Department of Information
Systems of the Faculty of Information Technology of BUT will help the European
Space Agency answer this question as part of the international research project
named Blockchain Enabled Deep Learning Data Analysis (or BLENDED in short). Apart
from the FIT BUT, the project participants include Belgian company
SpaceApplications, IT4Innovations National Supercomputer Centre in Ostrava, as
well as Greek partners Forth (Foundation for Research and Technology Hellas
research centre) and Geosystem Hellas (company specialised in processing of
geodetic data).

Together, the institutions will be working on solving one
of the long-term scientific projects of the European Space Agency which focuses
on the creation of a platform that will use machine learning for analysis of
space data. Over the years of its existence, ESA has used its satellites to
gather vast amounts of data and images of different places on Earth. This data is
freely available for universities and companies to subsequently analyse, process
and use to reach interesting conclusions - for example the rates of drying out of
soil, the rate of urbanisation or the fertility farmlands.

Processing of
these terabytes of data in real time is very difficult, therefore, artificial
intelligence is nowadays used to facilitate this task. It is quite easy to create
an algorithm solving a certain problem with respect to one specific dataset;
however, to adapt and successfully use such algorithm on thousands of different
datasets requires the use of AI. The research team from the Department of
Information Systems of the FIT BUT will take up the challenge of finding the best
way to share the results of such analyses, e.g. normalised data, extracted
photometric layers or trained AI models. Together, they have designed and are
implementing a platform that would make it possible to do just that. The platform
is based on two complementary technologies - InterPlanetary File System (IPFS)
and Ethereum. 

The NES@FIT research group participates in the project.
Members of the group are Vladimír Veselý, Dušan Kolář, Ondřej Lichtner, Michal
Koutenský, Dominika Regéciová and Matúš Múčka. They have vast experience with
cryptocurrencies, blockchain technology, smart contracts, and distributed systems
in general and they are developing a platform that has significantly larger
potential than just the required use within the ESA project. The platform can
work as a sort of undercarriage for any completely distributed system (in terms
of data storage, computing, and system management). So we are very much looking
forward to seeing how the know-how and experience acquired by the NES@FIT team
pay out in other grant opportunities or different contractual research
projects.

The co-operation with the research partners within the project
will continue until mid-2021 when the deployment of the platform prototype should
be completed which will enable the following:

• upload and (securely)
  share any data within a potentially unlimited storage;
• run series of
  highly demanding AI calculations (both third-party and the participants' own)
  using the data stored in datacentres participating in the
  project;
• subsequently publish (in the IPFS) the results of such
  calculations, algorithms used and the AI models trained, as data for which it is
  possible to