ForensicLMDQL

Authors:
Professor Paulo Caetano da Silva, PhD (Lattes CV)
Marcio Alexandre P. da Silva, M.Sc. (Lattes CV)

The current stage of Information Technology led many countries to adopt technologies based on international standards for the disclosure of their financial statements. Aiming to facilitate the exchange of data and increase the transparency of financial information available on the internet, languages derived from XML (e.g. XBRL) have been adopted as the standard technology for several companies. However, the occurrences of financial crime in large corporations - and also fraudulent electronic transactions - have generated losses of billions of dollars annually and have also attracted the attention of governments and regulatory agencies, fostering research for fraud detection using computational resources applicable to digital financial reports. With such goal, this dissertation proposes a tool for forensic financial analysts (i.e. forensic accountants or forensic inspectors) based on OLAP query languages, for the detection of fraud in financial reports represented in an the derivative of XML, i.e. XBRL, which is maintained by an international consortium composed of more than 600 companies and adopted by several governments. An extension of LMDQL language is proposed, which supports both queries on XML documents connected by XLink and XML Schema (intrinsic characteristic of XBRL), and on relational data. Thus, forensic operators which extend those defined in LMDQL are presented, and their use is demonstrated in a relational database, as well as in a native XML database. The relational data model chosen in this dissertation is based on XBRL 2.1 specification, which makes it independent of the business model of the organization that uses it, facilitating its application in different contexts. To evaluate the forensic operators, a case study was conducted from XBRL documents made available by the United States Securities and Exchange Commission (U.S. SEC). To load the relational repository an ETL processing is presented (Extract, Transform, Load) on financial reports, while in the native XML database the original format of XBRL documents was kept. Thus, it was possible to evaluate the efficiency of the proposal presented, in which probabilistic calculations used in forensic accounting were applied, on an OLAP server. An evaluation of the runtime was performed on the queries in relational and XML databases and it was verified that the processing of queries on the relational repository was faster.

Forensic LMDQL Operators (Files)

Published Scientific Papers:
1. EDOCW, 2014 IEEE 18th International: http://ieeexplore.ieee.org/xpl/articleDetails.jsp?reload=true&arnumber=6975384 (Ulm/Germany)
2. IADIS International Conference WWW/Internet 2014: http://www.iadisportal.org/digital-library/financial-forensic-analysis (Porto/Portugal)
3. Contecsi, 2013 10th International: http://www.tecsi.fea.usp.br/contecsi/index.php/contecsi/10contecsi/paper/view/223 (São Paulo/Brazil)

Master Dissertation (PT-BR) (pdf)