Network analysis is the key to understanding complex systems

Social networks, business processes, biological systems — all of this can be represented as nodes and connections. Network analysis reveals hidden patterns, and in fact-checking it helps to find sources of fakes.

Network analysis is an information analysis method that focuses on studying the relationships and structure in complex networks. This approach helps identify key participants, central nodes, and patterns in network relationships, which allows for a better understanding of the functioning of the system as a whole. Let’s take a closer look at the main elements of network analysis and its application in various fields.

Key elements of network analysis

Graph models. Network analysis is based on graph models, which are abstract structures consisting of nodes (vertices) and connections (edges) between them. These models help visualize and analyze complex relationships in data.
Centrality analysis: Network analysis involves the study of the centrality of nodes in a network. Centrality can be determined by various parameters such as degree (number of links), betweenness centrality (number of shortest paths through a node), closeness centrality (closeness to other nodes), and others.
Cluster analysis: This method is used to identify groups of nodes that are closely related to each other. Cluster analysis helps to reveal the structure of the network, identify communities, and highlight key groups in the data.
Modeling network dynamics. Network analysis can also include modeling network dynamics, which is the study of changes in the network structure over time. This allows one to predict the evolution of the network and identify possible changes in its structure.

Application of network analysis
One of the key applications of network analysis today is combating disinformation. Fact-checkers and analysts use the method to:

Identify sources of false information — using centrality analysis, you can determine which accounts or sites are most likely to spread fakes.

Example: In 2020, five popular Facebook pages spread fake quotes attributed to presidents and doctors. The posts fueled panic in the DR Congo and among the Congolese diaspora in France. Journalists used network analysis tools to track down the original sources. An analysis of the connections between the accounts revealed that a 20-year-old student and a 16-year-old schoolboy from Kinshasa were behind the campaign. They created fakes to attract followers. The exposure stopped the spread of dangerous myths.

A page that published fake posts about COVID-19 has gained 60,000 followers in one month.

Track the spread of fakes — graph models show how false news moves from social networks to messengers and the media.

Example: In 2022, a 2008 article about the “benefits of hunger” was taken out of context and distributed as “proof of the UN’s plans.” Network analysis helped identify the original source — an archived publication in the UN Chronicle — and trace the path of distortion — through the repost graph, they identified who first presented the material as a “sensation.”

Find coordinated campaigns — Cluster analysis helps detect botnets and coordinated groups that push false narratives en masse.

Example: In 2016 and 2020, on the eve of the US elections, residents of the Macedonian town of Veles created websites similar to news agency platforms, where they posted fake news about American politicians, and then distributed links to them using social networks. Network analysis helped to detect identical website templates, synchronous publications on social networks, and a common advertising network that brought in revenue for the authors from clicks. Thus, thanks to fact-checking and network analysis, an entire network was exposed that created fakes that influenced American public opinion and could potentially change the election results.

The creators of the network of fake sites deliberately made them look like portals of well-known news agencies.

In addition, network analysis is applied in the following areas:

Social networks: in sociology, network analysis is used to study social networks, identify key actors, groups and communities, analyze the influence and dissemination of information.
Business analysis: in business, network analysis is used to analyze business networks, identify key partners and customers, analyze competitive relationships and determine development strategies.
Biological networks: in biology, network analysis is used to study gene and protein networks, metabolic pathways and interactions in cells and organisms.
Information networks: in the field of information technology, network analysis is used to analyze Internet networks, computer networks, distributed systems and other information structures.

Challenges and prospects

Network analysis has its challenges, such as the difficulty of interpreting results, data heterogeneity, and the need to account for network dynamics. However, with advances in data analysis methods and computing capabilities, network analysis remains an important tool for understanding complex systems and the relationships between their elements.

Difficulty in interpreting results. Even after discovering a network of linked accounts, it is difficult to prove that this is a coordinated campaign and not an organic discussion. To reduce false positives, it is better to use additional metrics (temporal patterns, language markers).
Data heterogeneity. Social networks, forums, and messengers use different data formats, which complicates their joint analysis. The development of cross-platform tools helps to overcome this problem.
Network dynamics. Disinformation networks quickly adapt — they change tactics, create new accounts, migrate between platforms. To solve this problem, you can use real-time algorithms (for example, dynamic graph models).
The problem of scale. Social networks generate terabytes of data daily, making manual analysis impossible. Automation through machine learning simplifies the task.

However, the development of AI, predictive models and regulation open up new possibilities in the application of the method. The key trend is the transition from reactive denial to preventive threat detection. Prospects of network analysis:

Integration of AI and network analysis. GPT models develop capabilities for automatic analysis of text patterns in connections between accounts and search for reused media (for example, deepfake videos).
Predictive analytics. Thanks to network analysis tools, it is possible to model and predict the virality of fakes at early stages.

Network analysis is the “X-ray” of disinformation. It reveals hidden connections, helping fact-checkers not only find the sources of lies but also predict their spread, making it indispensable in the age of digital transformation.