Private Sector

Research Organisation

Coordinated between logos and skillsUniversity of Luxembourg (SnT, Interdisciplinary Centre For Security, Reliability and Trust), both teams built a bridge between private sector organizations and the world of applied university research.

Logos’ customers benefit a strong added values, especially in the context of new technologies using of Machine Learning, Artificial Intelligence and Federated Learning.

Name

Management

Using dedicated model is a proofed technical way to confirm if name of a physical person make sense compared to the context of analysis.
Innovation lies in the possibility offered to manage Arabic and Chinese names. The main advantage of our solution consists in learning and scoring the likelihood of a name (and its writing) for a given language.

User Profiling using

string similarities

Research done is to transform sequences of complex events into simple symbols and treat the user history a string. Then users can be compared using string similarities or other NPL techniques. The main advantage consist in having a simple way to compare behavior among multiple users, cluster user profiles and also compare an user with her/his past behavior.

Finding suspicious activities

in distributed ledgers

Analysis of very large set of big data exposed to several mathematical models has been completed to define the best scenarii and models to be used in cases of finding suspicious behavior in distributed ledgers, in order to detect anomalies and profile transactions . Approach is an ensemble of anomaly detection techniques that gives an anomaly score which can use for user rules or to show in GUI how suspicious it is.

Research has been published during the 2017 Institute of Electrical and Electronics Engineers (IEEE) conference on Data Mining Workshops ISSN number 2375-9259, Nov 18 New Orleans

Alternative methods for

Data Generation Modelling

Data generation is needed to deal with imbalance classification. Fraud causes are underrepresented and thus difficult to learn in a supervised learning method. Using specific networks, LOGOS and SnT demonstrate that these models proved their efficiency and allow to share models with 3rd parties to get additional data to train their models but without revealing the original private data. This methodology facilitates generation of a large set of complex data starting from a reduced set of real data and brings companies a high return on investment due to a huge cost cutting operation of man days required to validate business models on a large set of data.

Entity name identification

and relationships

Function allows to improve link analysis to different purposes; understanding relationships between different people, entities and topics for investigative purposes. Functionality automates reading and understanding of text paragraph and identify the relevance of a text document regarding the content combination of a (person) name and an associated topic, for example, to identify PEP (Political Exposed Persons) or any other legal person. (e.g. criminal behavior, asset ownership) Immediate business applications and benefits are that this technology could be used immediately in existing products to improve text search and provide more relevant results.

Graph based approach for suspicious behavior

detection in financial data

LOGOS and SnT contributions consist in modelling and discovering graph patterns to score them for detecting financial transactions suspicious behavior. The method is unsupervised approach. The financial expert automatically discovers patterns with their scores by modeling into a graph model. Graph model is used to periodically detect the suspicious patterns. The developed tool based on the graph-based models is fully automatic. The user (financial expert) only  fine-tuned the period to analyze the data. The time period could be a day, week, month or a year.  Thus, for each period, the abnormal transaction patterns are discovered.

ADF - Anomaly Detection Framework

Prototypes were organized in a framework to generate an aggregated score in order to concisely decide between normal and abnormal behavior. The proposed framework, Anomaly Detection Framework (ADF), is an assembled solution that combines miscellaneaous techniques: graph-based, clustering and decision threes.

Federated Learning

LOGOS and SEDAN team at SnT explore together a new paradigm for machine learning which is safe, secure and GDPR compliant, while still allowing multiple parties to collaborate and improve existing machine learning models. Federated Learning allows to train and synthesis multiple models without sharing any data. We are part of the project Coffe, lead by the SEDAN group at SnT, in which we validate our domain specific knowledge in finance, the academic and technological research.  Objective is to provide a common, robust machine learning model without sharing data, thus allowing to address critical issues for the financial sector such as AML/KYC/CFT and access to heterogeneous data.

R&D is part of our DNA

Partnerships

LUXINNOVATION

Is an organization that contributes to Luxembourg's economic development by stimulating innovation, fueling international growth and attracting foreign investors to the country. Until 2020, Luxinnovation has been a strategic partner for logos supporting our company in innovative activities.

ING Bank Luxembourg

InnovFin initiative

European Financing for Innovation is a joint initiative of the European Investment Bank Group (EIB and EIF) and the European Commission under the Horizon 2020 program. InnovFin aims to facilitate and accelerate access to finance for companies and other innovative organizations in Europe.

Luxembourg University

A partnership contract with the University of Luxembourg (SnT, Interdisciplinary Centre For Security, Reliability and Trust) allows our company to work with High-Qualified team of Researchers on complex subject like Detecting Suspicious Financial Transactions in Distributed Ledgers, Machine Learning, Federated Learning and Artificial intelligence in coordination with our R&D team. Results from applied Researches are integrated in its software.

DigitalEurope/APSI

DIGITALEUROPE represents the digital technology industry in Europe and APSI the digital industry in Luxembourg. Members include some of the world's largest IT, telecoms companies from every part of Europe. DIGITALEUROPE wants a European Union that nurtures and supports digital technology industries, and that prospers from the jobs provided, the innovation and economic benefits delivered and the societal challenges addressed.

Finance & Technology

Luxembourg

Since 2014, Logos is a member of FTL association which as mission to actively develop the Finance & Technology ecosystem in Luxembourg and abroad by informing, promoting, helping and leveraging synergies between its members, the Financial sector and other industries around the world. From 2014 to 2018, Logos got agreements 29.3 & 29.4 of Professional of the Financial Sector granted to IT companies by the CSSF, the Luxembourgish regulator. Since then, Logos kept its organization based on CSSF circulars.

E info@logos-its.com

T +352.26.36.55.1

GDPR Legal notice