DDoS attacks pose a significant threat to networks due to their ease of execution and the difficulty in mitigating them. Machine learning-based detection faces challenges due to limited accessible training data. This paper presents an instance-based transfer learning algorithm for DDoS detection, leveraging publicly available datasets for knowledge transfer. The approach filters and reconstructs instances using gain-based selection, TrAdaBoost, Q TRANSFER, and a Gaussian membership function to refine instance weighting. This enhances model generalizability to new attack types, ensuring robust detection. The proposed method improves adaptability to evolving threats, strengthening the reliability of DDoS detection technologies.