The Generals of Gold: A data-driven investigation of the Egyptian Army’s influence


Noonpost team members explain how they used Linkurious to conduct a large-scale investigation into Egypt's corruption networks.

The Generals of Gold is a project composed of 12 episodes where we present the intertwined relations of Egypt’s military generals and businessmen across kinship, affection, official positions, and benefit-based networks. Throughout the episodes, our network visualization and analytical articles show various conflict of interests and a systematic exploitation of the Egyptian economy favoring the ruling elites.

Our project began, when we asked ourselves why the 2011 Egyptian Revolution was unsuccessful. We came to realize that revolutionaries and activists did not really understand the complex nature of the struggle and the power of the Army’s connections.  So, we became more convinced about the importance of studying the Army’s complex networks of affiliates.

The software

Since complex networks and connected data are best presented in network and graph visualizations, we needed a software that excels at these types of analysis. In the end, chose Linkurious because of the following:

  • Our main requirement was to have an end-to-end solution that covers our research process starting from direct data entry to the graph database (Neo4j), to indexed instant-search, to visual exploration, all the way to running analytical queries. This is all supported by Linkurious.
  • Linkurious supports UTF8 (Arabic text in our case) in an excellent way. With a very minimal tweak, Arabic text search in Linkurious was enhanced by applying Arabic stemming, phonetics, tokenization, and other NLP-based rules.
  • We saw how Linkurious was used for ICIJ’s Panama Papers.
  • Linkurious’ great integration with Neo4j and the support of multiple data sources.
  • It was important for us to have a tool that supports pinning nodes and embed their coordinates when the visualization is exported to JSON. Linkurious supports this perfectly.
  • Linkurious’ instant geo-mapping.

Data gathering and modelling

We started by defining our research objectives, hypothesis, and the network boundaries. Then we created a comprehensive property graph model that can help us in answering our questions. After that, we started to collect the data manually. 90% of the data was collected from publicly available data (reviewing more than 1000 news sites and research studies, mostly state-owned media outlets). The remaining 10% was through private sources.

We used the Neo4j Graph Database to store and organize the connected data into a single data model, so that we can easily query the information to find patterns.

In the quest of understanding the complex nature of the state-sponsored corruption in Egypt, we decided to study three main entities - Person, Organization, and Project - and the key relations within and among them.


Image: Drawing connections between the three entities.

We decided to distinguish the formal position tie from the influence tie between Person and Organization, as well as Person and Project, in order to handle all the possible scenarios. Some government organizations have authorities and agencies that belong to them or affiliated with them, hence we created “Affiliated_with” tie to capture such a relation. Although Projects seem to be independent, it was important for us to capture those projects that were linked in a way or another, like being funded from the same IMF loan. Final note about the model is that we prepared it, and hence the collected data, to handle temporal/longitudinal analysis. This is why all the ties capture “start_date” and “end_date” of the relation between entities.

Visualization and exploration of 12 influence networks with Linkurious

We generated 12 networks where each network presented the complex relations belonging to a specific sector. Through these networks, we were able to validate one of the hypothesis we originally had which states that there is a competition between the Egyptian army and the Egyptian General Intelligence Directorate to dominate each sector, and that some private sector companies are paying the price.

Take for example the “Health Sector” network. Our investigation and analysis showed that the Egyptian army has a strong influence on Ministry of Health through two minister assistants: Hisham Abdelraouf (Primary Health Care Assistant) and Elsayed Shahed (Financial Assistant). Both used to have official positions in army-affiliated agencies. This has eventually led to rewarding the army’s agency “Armed Forces Medical Services” with key projects to the Egyptian economy.


Image: Links between the minister assistants and the army’s agencies leading to key projects.

In most projects, the armed forces contracts with private sector companies, which are connected in a way or another to the army. As an example, importing the children/infants milk project was given to Pharma Overseas Company whose Chairman of Board of Directors is Ahmad Jazzarin, the brother-in-law of Vice Admiral Mohab Mamish, chairman of the Suez Canal Authority and former Commander of the Egyptian Navy.


Image: Pharma Overseas Company links to the army through direct work on a key project as well as kinship ties to Vice Admiral Mohab Mamish.

Moving to the Egyptian General Intelligence Directorate (EGID), our analysis has shown that there are two key private sector companies linked to EGID that handle large scale projects in the health sector: The Egyptian Pharmaceutical Trading Company and Wadi El Neel For Medical Services.


Image: EGID’s affiliated companies and how it influences MoH.

Similar findings were created by studying the “petrochemicals, energy and minerals“, “communications“, “industry“, “tourism”, “food”, and “national projects” sectors.

Running queries to identify patterns

Generally speaking, the network boundaries, collected data, and research questions that we defined led us to rely heavily on visual exploration. However, we used various queries that helped us to identify interesting starting points and patterns. For example we used this query to find people who have kinship tie with others who own at least one private company:


Represented visually as:


Lessons learnt and next steps

Whether is it to investigate financial ties, leaked documents or to understand and highlight influence networks, our project showed us that graph technologies are a powerful tool to find stories hidden in large data sets. To this end, Linkurious allowed us to visually explore and query highly connected data in order to reveal the sometimes not-obvious relationships between entities.

Over the course of the project, we identified three important lessons:

  • The importance of defining clear research objectives and practical network boundaries.
  • Defining robust data model is very essential step to answer the questions and test hypothesis.
  • Applying continuous exploration and visual analytics is always important to identify gaps in data and to extract key insights.
  • The next step is to enhance the model and data we have to build temporal/longitudinal networks, to understand the rise and fall of specific networks communities in chronological order. Moreover, we will keep updating our data and explore the possibilities of enriching our data with publicly available data sources in an automated way.

Explore Generals of Gold here.