Gdansk

Jana Haj Darwish - Portfolio - Gdansk 25-26.pdf

#Why

#History

#Concept

#Data Sets

#physical Bastion

#Digital Bastion

A 14-Year View of Poland’s Cybersecurity Landscape Through SOM Clustering Self-Organising Map (SOM) generated from 14 years of CERT Polska annual reports. Each dot represents a keyword extracted from the reports, and the clusters reveal how the focus of cybersecurity topics evolved over more than a decade.

1. Collecting 14 Years of CERT Polska Reports 2. Extracting Text and Cleaning the Data 3. Building a Keyword Frequency Matrix 4. Training the Self-Organising Map (SOM) 5. Visualising Each Year’s Landscape 6. Identifying Long-Term Patterns

Key SOM parameters:

Neighbouring function: Gaussian Learning rate: 0.5 Sigma: 1.0 Training iterations: 3000 randomised updates Initialisation: Random weight initialisation with normalisation

Cluster assignment: Each site was assigned a cluster ID based on its Best Matching Unit (BMU) coordinates. Clusters were analysed for common patterns, such as historical periods, dominant criteria, or geographic concentration.

U-Matrix (Unified Distance Matrix): Shows the distances between neighbouring neurons. Light regions represent homogeneous clusters, whereas dark boundaries indicate sharp changes in site characteristics.

Hit Map: Displays how many sites map to each SOM neuron, revealing dense cluster areas and sparsely populated regions.

Component planes: Individual feature maps for each variable (e.g., century, criteria v), highlighting how attributes contribute to cluster formation.

Geographical visualisation: Cluster groups were projected onto a world map to examine spatial coherence and identify regional defensive traditions.

Scrape a timeline of weapons from: https://worldostats.com/timelines/evolution-of-warfare/#google_vignette

Visualise weapon evolution over time using Plotly scatter plots.

Interactive chart:

Then, I improved the chart dividing the Y axis by weapon type to get better data set and improve the outcome:

warfare_timeline_log_axis_with_region.html Interactive chart

for the same data set,

Scrap 14 Cyber attacks reports from : https://cert.pl/en/publications/

Get information from the PDF to txt

Extracting data from text

Loading the dataset and visualising it using Matplotlib and Seaborn

Link to Script files: unesco_defensive_sites

List of bastion forts - Wikipedia Script Files : List of Bastion forts-WIKI