We consider issue-based computer platforms, designed to transfer knowledge in a way, that ensures its thorough and effective utilization in various domains. Such platforms have subsystems for collecting knowledge and for decision-making support. Typically, these systems are centrally located and store data in one specific place, such as a data center, or even in one database, where this data can be easily lost or damaged. In addition, this data can be spoofed or altered anonymously in many ways, if traditional software or hardware access to it is attacked or misused. These problems can be partially solved by introduction of proper security and monitoring solutions, using the best security practices, such as secure protocols and first-class encryption. However, this article offers methods that, practically, eliminate the very possibility of data falsification, and add other important properties to computer systems, such as data invariability, decentralization, and fault tolerance of the data registry or individual subsystems of issue-based knowledge transfer platform.
There is a global dilemma of the most thorough use of available knowledge. The
effectiveness of applying this knowledge to various subject domains is one of the most
important factors contributing to overall human progress. According to the research,
conducted by a US company Delphi Group at the beginning of the current millennium [
There is no doubt that when making decisions in different fields of activity, it is necessary to rely on all available knowledge, both stored somewhere and provided by an expert ad hoc. This is especially true for loosely structured data, characterized by high level of uncertainty and incompleteness.
The first stage of solving of the above-mentioned problem could be integration of knowledge of different formats, obtained from different time-dependent sources, with formalized knowledge and knowledge that experts provide. If combined with processing and storage of this formalized data in databases, this approach can become an ultimate solution. Each next transfer of accumulated relevant knowledge for its further use in decision-making process can be the final stage of decision-making, which brings the desired results. An example of the implementation of this approach is the issuebased hardware and software platform of knowledge transfer, created according to the chart, shown outlined on Figure 1.
Functionally, the knowledge transfer platform is represented by the following
components:
1. The subsystem of knowledge collection and actualization, which is designed to
support the experts’ workflow, knowledge engineers (expert session organizers) and
decision-makers (DMs). The main functions of this subsystem are:
user registration and accounting.
assessment of competence for engineers and experts, which provides insights on
what knowledge fields they are most experienced in [
It is assumed that the knowledge engineer, who is also the main actor in the system
and who has knowledge in specific subject domains, is using these tools for automated
extraction of knowledge and for organizing expert sessions. Engineers, using DSS
tools, create models of subject domains and knowledge databases. Later, based on these
models and knowledge, decision makers get recommendations from the system.
Functionally, in the knowledge transfer platform, the decision support system (DSS) is
responsible for generating recommendations [
In this paper, decentralization means the use of a cryptographically protected decentralized data platform, which works based on blockchain technology. Decentralizing a centralized system is an important process, which can also be considered a mandatory one, if there is a need to ensure timely and ultimate data protection against unauthorized interference and targeted attacks. Before blockchain was invented, this task was solved using multiple secure layers and geo-decentralization, which, however, could not guarantee absolute data protection.
System decentralization process is the transition of the system from fully centralized
to partially or fully decentralized mode. Decentralization makes sense if we need the
system to assume the following properties [
When it comes to decentralization of systems built with decentralized data platforms,
the following types of decentralization should be outlined:
Full decentralization. Systems of this type assume that their computational part,
including the data management and storage subsystems, is also decentralized and,
therefore, cryptographically protected. This is only possible when the program is
fully stored in a decentralized data platform, and no parts of the system are outside
that platform. Modern decentralized data platforms, such as Ethereum and similar
ones, allow you to create the so-called decentralized programs that are loaded into a
decentralized data registry only once and can never be changed again. This
immutability gives strong security advantages, but it means that the software should foresee
all possible scenarios of interacting with it in advance, because, after being loaded,
it will never ever be changed. Thus, there will be no practical possibility of data
forgery and any other methods of interference with the originally uploaded program.
This also requires the program to be carefully reviewed for any vulnerabilities before
uploading it to the decentralized registry [
The level of decentralization of any system is determined not only by how it uses
decentralized data platforms and integrates with it, but also by decentralization level of
the decentralized data platform itself (for instance, which algorithm it uses to protect
data and computation centralization). However, this problem is out of scope of this
article [
When it comes to distributed systems for collecting expert knowledge, which can be
related to partially decentralized systems, the knowledge transfer platform acquires the
following important properties:
Guaranteed reliability of expert knowledge storage, including permanent availability
of this storage for a small (but sufficient) amount of data needed for DSS system,
and complete protection against unauthorized changes. This property is especially
important for expert systems and knowledge-based systems, because often the
results of a certain examination, after its completion, must be stored for a long time,
for example, up to 10 years from the moment of completion of some research
projects.
Possibility of conducting a public audit of the system later, based on reliable input
data, stored in the decentralized registry. Since the data, stored in the decentralized
registry, is cryptographically protected, there is no doubt in its authenticity. Using
this reliable data, assuming that the DSS is idempotent when performing calculations
(it produces the same result with the same input data), it is possible to verify the
result and the reliability of the source data by performing the calculation again.
As a result, we have increased confidence in the functioning of the knowledge
transfer platform, which also increases the credibility of results, obtained from the DSS.
An alternative to decentralization of the subsystem for collecting expert knowledge of
an issue-based knowledge platform is the use of electronic digital signatures (EDS) and
several centralized data registers (like traditional databases). It is worth noting that even
in the case when these alternatives are used, the knowledge collection system still won’t
have the properties, listed above [
A recently conducted research demonstrates, that in a typical system of expert knowledge collection, the total amount of data (which primarily consists of expert estimates) recorded in the decentralized register is small, and occupies less than 500 kilobytes of storage space per project. Since the results obtained and stored by the DSS may be in demand (reviewed and used) for quite a long time after the project completion (in extreme cases, up to 10 years), we conclude, that for such systems it makes sense to store all input information in a decentralized register. In other words, a complete decentralization of a system, which collects expert knowledge, means partial decentralization of the knowledge transfer platform itself, since the system of expert knowledge collection is its subsystem.
However, full decentralization of a system always makes it somewhat difficult for
the user to work with it. After decentralization it is necessary to have not only a
decentralized account (also known as a "wallet" in cryptographic decentralized data
platforms), but also a certain number of cryptocurrencies to pay for transactions in the
decentralized network. This problem is classified as a problem of adoption of technology
and is one of the reasons why many systems and products cannot integrate with
decentralized technologies just yet [
After decentralization, the system, which collects expert knowledge, will not have fundamental differences from the previous version of the system, except for several additional prerequisites the experts should do before working with the system. There are also a few steps, which knowledge engineers should also do, but only once before starting any project: ─ Creating a decentralized program which will be uploaded to the decentralized data platform. This program can be templated, that is, developed once, and only slightly modified for each specific subject domain using pre-programmed parameters. ─ Creating a delegate account and providing it with the cryptocurrency, necessary for further decentralized transactions which experts will perform. ─ Experts, in turn, do several additional simple steps: before entering their estimates into the system, each expert generates his(her) own decentralized account, and uses the graphical user interface to register it in a decentralized program (separately for each expert session project or only once, depending on the system design), thereby authenticating in the system, and confirming his(her) identity. Experts can associate any additional data with this identity. No one but experts themselves can input evaluations on their behalf after such registration in a decentralized system.
The preparation work consists of the following steps: 1. The administrator develops (or clones an existing) decentralized program created for a specific expert system (or uses a template). The function of this program is to perform a cryptographically protected entry of the expert data into a decentralized register with the help of a delegate, for further reading by the decision support system. 2. The administrator registers this program in a decentralized data platform, thereby making it immutable. It obtains a unique application address, which is also immutable. 3. The received address of the decentralized program is written into the centralized storage (or program code) of the expert system for further interaction with it. 4. Additionally, the administrator generates a delegate account and replenishes its balance in a decentralized system. The delegate account is also stored in the delegated transaction support system, which can a be part of the decision support system. 5. Before starting to work with the system, experts create their decentralized accounts (for example, using the browser extension called Metamask) and register them in the centralized system by creating electronic digital signatures for certain pieces of data, requested by the system (for example, an email address). Thus, they confirm that they are the owners of this decentralized account. 6. After the work with the system is done, the expert performs a delegated transaction to write the generated data into the decentralized storage. This transaction is genuine, and no one but the expert himself can change this data after it is stored in the decentralized registry, and, moreover, at any time anyone can check the correctness of the data, that was written into the decentralized network (although this is not necessary).
While expert knowledge is processed by the decision support system, the data is read from a decentralized registry and converted into a format, that is convenient for processing. In future, data from the decentralized registry can also be used, for example, to audit a certain expertise (session). Figure 3 shows the interaction of an expert and the centralized expert data collection system with the decentralized program before and after the examination (i. e., how experts work and read data from a decentralized register).
After decentralization, when the changes, suggested in the paper, were applied, the (previously centralized) system of expert knowledge collection, acquired the following properties: 1. The input data, obtained from experts, is now protected and maintained by a global decentralized data platform, which guarantees its storage and immutability. 2. Since the data is cryptographically secure and immutable, it is considered a reliable source of credible knowledge (provided that the decentralized program is correct, which can be easily verified at any time). This allows us to increase the level of trust to the decision support system itself, and to conduct a post factum auditing of the system, if necessary.
In order to implement the above-listed properties within the system, only minor changes had to be introduced into the existing centralized system.
Here they are: 1. The need for experts to install an additional (relatively simple) software, as either a mobile application or a browser plugin. This approach, in its turn, can also completely replace the process of registration in the expert data collection system using traditional login and a password. 2. Adding two new steps for the expert to perform: electronic data signature before (registration) and after (executing a data recording transaction in a decentralized register) working with the centralized system.
The result of the conducted research is the method proposal for decentralization of the issue-based knowledge transfer platforms, which include subsystems for knowledge collection and decision-making support. The proposed improvements can significantly increase the security of information storage and processing in the system, allowing it to prevent any loss, as well as unauthorized storage-level changes of information and changes to the original software.
As a result of decentralization, the data, obtained from experts, is protected and maintained by a global decentralized data platform, which guarantees its lifetime storage and inability of intruders to compromise it. Once recorded, cryptographically protected data remains unchanged and can be considered a reliable source of information, also verifying that if the centralized software is correct and unchanged. The suggested approach completely eliminates the possibility of data falsification and adds important properties to the system, such as invariability and fault tolerance. As a result, decentralization allows to increase the level of confidence in the knowledge transfer platform and perform open system audits when necessary.
This work has been produced in collaboration with the European Union’s Horizon 2020 research and innovation programme under the grant agreement N 830943, ECHO project.