Security is critical for reliable service delivery and ensuring overall business continuity, whether it is smart manufacturing, healthcare, cyber-supply chain, or any other infrastructure using a digital ecosystem. A business model is itself an environment that consists of various complex processes, devices, users, developers, and systems. Every constituent of this ecosystem is equally valuable to the business, and a threat posed on its entities poses a greater risk to the business and its owners. Ensuring cybersecurity within modern business ecosystems is essential to maintaining reliable service delivery and business continuity. Today, smart cyberattackers are unavoidable, but early prediction of such attacks will help organisations to prepare for an attack or before such an attack. Thus, this paper examines the role of predictive analytics and machine learning models in identifying and predicting cyberattack patterns to secure these digital ecosystems. CRISP framework integrates predictive analytics with multiple machine learning models (Random Forest, Logistic Regression, XGBoost, CatBoost, LightGBM, CNN, LSTM, GRU, and ensemble approaches) to enable proactive cyber resilience. In experiments conducted on the Microsoft Malware Prediction dataset with over 2 million samples and 54 refined features, CatBoost achieved the highest AUC-ROC score of 0.723, while ensemble methods achieved an accuracy of 65.8% and an AUC of 0.721. These results demonstrate the robustness and diversity of the evaluated models, confirming the effectiveness of our proposed system in predicting and prioritizing threats. Additionally, we introduce a threat categorization system based on urgency levels, enabling businesses to prioritize their defensive measures effectively. Our approach aims to provide businesses with actionable insights for robust cyber-resilient systems, emphasizing the importance of preemptive threat management in the face of evolving cyber threats.
Verma et al. (Thu,) studied this question.