A Data Enhancement Model for Intrusion Detection In Smart Home

Smart home devices differ in data throughput and attack frequency. However, some devices lack sufficient attack data, resulting in small sample sizes that pose a challenge to training intrusion detection models. Therefore, this paper proposes a data enhancement model DE-CWGAN (Data Enhancement Based on Conditional Wasserstein Generative Adversarial Networks) to generate effective training samples for small sample data. The DE-CWGAN model uses CWGAN as a method for data enhancement. The sampled data are used as real data samples to train the CWGAN, and the generator is used to produce generated data samples (fake data). Wasserstein distance is used to calculate the loss between fake and real data, and the generator is adjusted based on this loss to generate fake data i.e closer to the real data. Add the fake data to the original dataset to expand the dataset. The experimental results show that the DE-CWGAN model effectively fills sparse attack records, and the machine learning classification algorithm achieve improved accuracy after training with enhanced data.