Downloads: 0
Research Paper | Computer Science and Engineering | Volume 15 Issue 4, April 2026 | Pages: 195 - 198 | India
Forecasting Cyber Threat Trends: A Hybrid Statistical-Deep Learning Approach
Abstract: The fast growth of technology has resulted in significant rise in cyber-attacks. It creates substantial challenges for organizations and governments worldwide. Traditional cybersecurity measures are not sufficient to address growing attack patterns. This study presents a forecasting framework using a Hybrid Statistical Model ARIMA and Deep Learning Model LSTM to predict year wise cyber-attack trends and number of affected users across different attack types. The linear component of attack pattern is captured by ARIMA and the nonlinear residual pattern are modeled using LSTM networks. This combination improves the prediction accuracy to 100%. Historical Time Series cyber-attack data from 2015-2024 is used for training and evaluation. The hybrid model gives higher performance with MSE, RMSE and MAPE validating its effectiveness. The framework provides forecasting of next five years for actionable insights for cybersecurity planning, resource allocation and risk mitigation.
Keywords: Cybersecurity, ARIMA, LSTM, Time Series Prediction, Cyber-attacks
How to Cite?: Jayashree P. Darade, Dr. Anita Chaware, "Forecasting Cyber Threat Trends: A Hybrid Statistical-Deep Learning Approach ", Volume 15 Issue 4, April 2026, International Journal of Science and Research (IJSR), Pages: 195-198, https://www.ijsr.net/getabstract.php?paperid=SC26211111417, DOI: https://dx.dx.doi.org/10.21275/SC26211111417