As the sophistication and frequency of cyber threats increase, there is a pressing need for adaptive, scalable, and privacy-preserving Intrusion Detection Systems (IDS) to secure diverse and interconnected network architectures. This paper presents an advanced IDS framework that integrates Federated Transfer Learning (FTL) with Differential Privacy (DP) to address these challenges. FTL enables distributed networks to collaboratively train a global model that learns from individual networks while keeping data decentralized. In tandem, DP in Federated Averaging (FedAvg) preserves each client’s data privacy by adding controlled noise, ensuring that sensitive information is protected from potential data reconstruction attacks. The proposed framework, Aegis, is designed for improved IDS performance, real-time adaptability, and privacy resilience in the face of evolving cybersecurity demands.