Secure by Design: Applying AI to Automate Threat Detection and De-Identification in Enterprise Systems

Abstract

The escalating complexity of cyber threats necessitates the adoption of secure-by-design methodologies in enterprise systems. This study examines the role of artificial intelligence (AI) in automating threat detection and enhancing de- identification processes to bolster resilience and privacy in enterprise settings. Grounded in adaptive security and privacy- preserving computation theories, this study utilises machine learning and deep learning models for intrusion detection, alongside AI-enhanced pseudonymization and generalisation techniques for de-identification. The findings indicate that AI- driven detection achieves accuracy rates exceeding 90%, surpassing traditional rule-based systems in identifying novel and evolving threats. Furthermore, AI-enhanced de-identification effectively balances privacy and utility, enabling enterprises to comply with regulatory mandates, such as the GDPR and HIPAA, without compromising data usability. Key challenges, including computational overhead, explainability, and adversarial resilience, were identified however, modular architectures and GPU acceleration mitigated the integration barriers. The study concludes that AI operationalises the secure-by-design paradigm by addressing the enduring trade-offs between privacy, security, and efficiency. Future research should investigate explainable AI, adversarially robust privacy methods, and quantum-safe architectures to ensure sustainable protection in an evolving threat landscape.