Analýza a mitigace phishingových útoků - případová studie

Phishing attacks currently constitute one of the most widespread and severe cyber threats to organizations in both the public and private sectors. The advent of artificial intelligence technologies, particularly Large Language Models (LLMs) and deepfake tools, significantly enhances the sophistication and personalization of these attacks, rendering traditional technical defense measures inadequate without appropriate employee training. The primary objective of this master's thesis is the empirical analysis of the effectiveness of simulated phishing campaigns conducted within a Prague-based joint-stock company using the Microsoft Defender for Office 365 platform. The research was carried out as a single-organization case study comprising three measurement rounds (R1–R3) and two training interventions. It involved a baseline susceptibility measurement of 329 employees, targeted embedded training via the MS Defender for Office 365 Attack Simulation Training platform, and a subsequent comprehensive retest of 378 employees four months after the initial intervention. The results demonstrated a statistically significant decrease in the click rate on phishing emails following the completion of the embedded training and subsequent in-person sessions, as well as the persistence of this effect over a four-month period. Consistent with existing literature, an interactive, behaviorally-oriented approach to training appears more promising compared to the traditional, strictly lecture-based format. Furthermore, demographic and organizational factors predicting higher susceptibility were identified. Based on these findings, a replicable methodological framework was developed for the regular testing of cyber resilience in municipal organizations, alongside a proposed set of specific recommendations for configuring training programs and security policies.