Search Results (1)

Despite the exponential rise in cybersecurity incidents worldwide, existing analytical approaches often fail to detect subtle temporal dynamics in cyber threats, particularly on a quarterly scale. This paper addresses a critical research gap in the domain of temporal cyber risk analysis by introducing a mathematically rigorous and AI-augmented framework capable of identifying, validating, and forecasting quarterly shifts in global cyber-attack patterns. The methodology integrates a hybrid data acquisition pipeline with GPT-based AI classification to construct a structured, high-dimensional dataset comprising 11,497 cybersecurity incidents spanning from October 2023 to March 2025. These incidents cover 106 attack types, 29 industries, and 257 countries. The framework decomposes the dataset into quarterly intervals and applies mathematical formulations to compute frequency shifts across categorical variables (attack types, industries, countries) and numerical variables (attack significance), followed by robust statistical validations (Chi-square and ANOVA tests), time-series forecasting via ARIMA, and the computation of a Quarterly Composite Index (QCI). Key results reveal dominant attack types—Social Engineering (ing1733) and Zero-Day Exploits (1657)—and highlight sectoral vulnerabilities in IT (5959) and Government (2508). Statistically significant quarterly variations were confirmed (${\chi }^2=2319.13$, $F=3.78$, $p<0.001$). ARIMA forecasts predict 1782–2080 incidents per quarter for 2025–2026, while QCI trends average around 0.75, signifying sustained volatility. The research delivers both theoretical and practical advancements by combining generative AI, temporal segmentation, and statistical modeling to create an operationalizable intelligence system. This contribution enhances strategic cybersecurity preparedness and policymaking in a complex, evolving threat landscape. Full article