KnowledgeX: Trusted data-driven knowledge extraction

ABOUT THE PROJECT

KnowledgeX: Trusted data-driven knowledge extraction

Over the past years, value generation for many businesses has become more and more data driven. Energy companies can reduce their CO2 footprint by analysing their operations, logistics companies can shorten their supply chain by optimizing their processes, and healthcare service providers can improve their patients' lives by data-driven prognostics.

Creating knowledge from data often requires highly specialized data scientists with deep domain knowledge and expertise in data analysis and, lately, especially machine learning. Nevertheless, the process of transferring highly valuable data to a data scientist for knowledge generation is privacy- and trust-intensive. Data providers want to be sure that the data scientist has the experience needed and only handles the data according to their authorization.

Therefore, KnowledgeX proposes a decentralized platform that (a) lets companies find specialized data scientists for their specific use cases in a marketplace, (b) enables traceable and transparent knowledge generation with trusted execution environments and blockchain technologies, and (c) maintains decentralized reputation storage of data analysis and knowledge generation activities carried out by data scientists. KnowledgeX aims to develop an ONTOCHAIN off-chain knowledge management application that acts as a catalyst for highly privacy- and trust-intensive knowledge generation for businesses, citizens, and communities built on top of blockchain and the next generation internet.

Motivation for the project:

Data Science is currently a trust-intensive process. Giving data scientists the full control of data is a risk. KnowledgeX enables data scientists to do their jobs while not giving away full control.

Generic use case description:

KnowledgeX enables trust-aware knowledge extraction from data without giving away control of it. Therefore, KnowledgeX uses blockchain and trusted execution environments.

Essential functionalities:

Matchmaking between data owners and data scientists, trusted execution of data analysis.

How these functionalities can be integrated within the software ecosystem:

KnowledgeX is applicable to any situation where knowledge for a specific problem is needed and data is valuable.

Gap being addressed:

Trust-aware data science.

Expected benefits achieved with the novel technology building blocks:

KnowledgeX enables a new paradigm of knowledge generation. It enables outsourcing highly trust-intense tasks to qualified persons without a need for complex NDAs.

Potential demonstration scenario:

Data science for IoT time series analysis and prediction.

PROJECT OUTCOMES

A company can work privately on their data with an expert data scientist without surrendering full control over the data.

Demo:

Repositories & Documentation:

www.knowledgex.eu

More details:

Customer engagement

Our application has a web interface through which the two customer segments can interact.

Monetization

Data owners pay the data scientists for their services. KnolwedgeX gets a cut of that fee.

Scenario

Step 1: Create a gig

Step 2: Find a data scientist

Step 3: Establish an agreement

Step 4: create data processing code based on sample data set

Step 5: execute code on full data set in TEE

Semantic content and content transfer

New knowledge is created in every interaction. Mapping the skills of data scientists with the requirements for a job are classified according to an ontology.

Ownership

The interactions and the new knowledge generated in the gigs are owned by the data owners.

Existing similar solutions/services

There are single generalized freelancer marketplaces (upwork etc.) that do not provide any privacy to the data being processed.

TESTIMONIAL

We enjoyed being part of ONTOCHAIN, especially the work with Luis Carlos and Anthony from iExec as well as Alberto and Marco from Intellisemantic. Overall, the work was very productive and we achieved a lot. However, the project had for us a lot of overhead. Four deliverables with on average in our case 70 pages within 6 months is a lot and we recommend for future executions to cut this overhead down to at max. 2 deliverables.