Owing to the rarity of fraud in non-life insurance sector, in this paper, we propose a Chaotic Variational autoencoder (C-VAE) to perform one class classification (OCC). Here, we employed the logistic chaotic map to generate random noise in the latent space. The effectiveness of C-VAE is demonstrated in the health insurance fraud and auto insurance datasets. We considered vanilla VAE as the baseline where Multi-layer perceptron (MLP) is employed in both Encoder and decoder architectures. Additionally, we proposed C-VAE-Wavelet Neural Network (C-VAE-WNN) too. We conducted sensitivity analysis by varying the latent dimensions and the number of hidden layers. It is observed that most of the time, C-VAE-WNN outperformed VAE in both datasets and complexity. Further, the t-test conducted inferred that C-VAE-WNN is statistically significant in both Medicare Part B and Automobile insurance datasets. In addition, we empirically demonstrated the statistical robustness of variants of C-VAE i.e., C-VAE-MLP and C-VAE-WNN by using the coefficient of variation.