Network intrusion and attack detection is an important and difficult task in the community of network security. Many researchers have conducted a lot of efforts and proposed many anomaly detection methods for network intrusion detection. The methods include data mining based anomaly detection (Wang 2018), fractal time series based anomaly detection (Radivilova 2019), information fusion based anomaly detection (Zhang 2008), principal component analysis based anomaly detection (Salman 2018), wavelet analysis based anomaly detection (Lu 2009), and fractal feature parameters based anomaly detection (Ya-min 2009). These methods performs feature analysis from different aspects to establish anomaly detection model and have been well applied in practice. These methods focus on how to extract features to train anomaly detection model, which can achieve a high detection accuracy. However, it requires massive labeled samples which are difficult to collect in actual network environment.