Brain is the Mopso offering for AML schema discovery; it integrates and analyzes information from multiple data sources using AI algorithms to speed up the AML processes.

The Mopso Brain is a modular software platform that runs on customer premises in a Kubernetes infrastructure. Cloud installation is also possible.

The Brain platform is part of the Mopso architecture; it describes financial facts with statement expressed using the Mopso Language Profile within the W3C Resource Description Framework (RDF). It also stores all RDF statements and their related meta-data in a knowledge graph .

Mopso Brain learns first-party data from the customer information systems; it enriches them with the second-party and third-party data and OSINT interacting with the Mopso Net service.

The Mopso Brain platform provides some packages containing specialized software agents that navigate the knowledge graph, infer data, find suspicious AML patterns, display relevant data, compute risks, enrich data, etc., etc.

The following picture summarizes the overall architecture of Mopso Brain :

brain architecture

The core of Mopso Brain is the KB component (in green) that is based on a optimized and customized version of the LinkedData.Center SDaaS platform; it provides:

  • a persistent graph database with lexical index and geolocation computation
  • a SPARQL endpoint fully compliant with the 1.1 version of UPDATE and QUERY W3C specifications
  • an integrated ASA engine embedding a reasoner and an ingestion agent
  • a web workbench to access all component features from the web interface (suitable for admins)
  • embedded support for Mopso Language profile

The API component (in yellow) provides a set of named parametric SPARQ queries supporting content negotiation. The queries are available through a REST interface described with Swagger. A web interface allows browsing the APIs catalog.

On the top of the SPARQL and REST APIs, there are some UI components (in blue):

  • a semantic search engine (google like)
  • a semantic browser aligned with semantic web standard
  • a pluggable tool for social network analysis

The data ingestion requires a legacy proxy component that provides linked data to the KB described with the Mopso Language Profile. ETL processing takes place on the legacy proxy component or in the autonomous agent component that also has asynchronous enrichment capability of the data in the knowledge base through Mopso Net services

Requirements

The system requirements for the Mopso Brain platform depend on the complexity and the quantity of information managed. The minimum requirements are:

  • a Kubernetes cluster version 1.19+ containing at least one node
  • a minimum of 6GB RAM available to the Mopso Brain core components (10 suggested). Add-on packages may require additional resources.

Note that Mopso Brain, by itself, does not provide network access to its components outside the Kubernetes cluster. The customer is responsible for configuring its cluster, namespaces, ingress services, load balancers, and API gateways according to its security policies.

Mopso specialists are available to support the customer and/or the system integrator in configuring a production system.

Available Packages

Core package

The core packages consist of the kb system component that realizes a knowledge base that must be populated with facts extracted from linked data expressed using the Mopso [language profile]. Some reasoners materialize inferences in the background, and some other reasoners remove stale data. The core package provides a SPARQL interface and some REST APIs to introspect the knowledge API.

component Type package description
brain/api-core Rest API brain-core REST APIs for the core package. Can be extended for special purposes
brain/kb Knowledge Graph brain-core implementazione custom di Graph database as a Service per brain

The core package requires to be connected to the legacy proxy component that provides the linked data resources to be ingested.

This picture summarizes the core package content:

core-package

Note that the core-package by itself does not provide any results. It SHOULD be attached to an ingestion agent that provides the facts to the knowledge base or to an autonomous agent providing data through the lp component.

UI package

The UI package consists of two components that interact with the core package:

component Type package description
brain/sse Web app brain-ui Semantic search engine for brain knowledge bae
brain/sb Web app brain-ui Semantic search browsing

This picture summarizes the UI package content:

ui-package

UI components MUST be accessed from an authenticated proxy managed by the customer.

Reference deploy schema

The Mopso Brain system is composed of a set of orchestrated docker containers generally including both the core and UI packages (in green) and some custom components that provide data enrichment, facilities, and access (in white)

All system components are designed to run under the control of a Kubernetes cluster or locally using docker-compose. Here is a typical deploy schema that you can use as a reference.

deploy

This deployment of Mopso Brain MAY implements the following process:

  1. Customer provides raw data in a shared volume as defined during the analysis phase.
  2. The autonomous agent (aa) executes ETL and enrichment tasks: it generates linked data files inside the data volume.
  3. The Mopso Brain, through the legacy proxy (lp), detects changes in the linked data learning all changed facts and populating the knowledge base
  4. Mopso Brain reasoner, in the background, construct the required inferences;
  5. the aa component search the Knowledge Base for eventually missing facts and, using the Mopso Net Services,
  6. the aa component looks for enriched information cooperating with the Mopso Net Services and puts the enrichment data in the shared volume;
  7. The API component highlights suspicious situations and missing data.
  8. The semantic search engine (sse) and the semantic browser components provide the customer with tools to visualize, search, and analyze the knowledge base.