G2D: Generate to Detect Anomaly

Abstract

In this paper, we propose a novel method for irregularity detection. Previous researches solve this problem as a One-Class Classification (OCC) task where they train a reference model on all of the available samples. Then, they consider a test sample as an anomaly if it has a diversion from the reference model. Generative Adversarial Networks (GANs) have achieved the most promising results for OCC while implementing and training such networks, especially for the OCC task, is a cumbersome and computationally expensive procedure. To cope with the mentioned challenges, we present a simple but effective method to solve the irregularity detection as a binary classification task in order to make the implementation easier along with improving the detection performance. We learn two deep neural networks (generator and discriminator) in a GAN-style setting on merely the normal samples. During training, the generator gradually becomes an expert to generate samples which are similar to the normal ones. In the training phase, when the generator fails to produce normal data (in the early stages of learning and also prior to the complete convergence), it can be considered as an irregularity generator. In this way, we simultaneously generate the irregular samples. Afterward, we train a binary classifier on the generated anomalous samples along with the normal instances in order to be capable of detecting irregularities. The proposed framework applies to different related applications of outlier and anomaly detection in images and videos, respectively. The results confirm that our proposed method is superior to the baseline and state-of-the-art solutions.

Maximising robustness and diversity for improving the deep neural network safety

“…However, their vulnerability against adversarial samples makes them insecure for critical tasks. They are prone to a class of attacks conducted in order to sabotage their performance referred to as adversarial attacks, which could cause missclassification or confidence reduction in sensitive tasks such as malware/anomaly detection [1,2] or autonomous driving [3] hence, a measure to face these attacks is of utmost importance. Generally, three strategies are presented: (1) adversarial training [4], (2) defining a robust loss for learning CNNs [5], and (3) refining [6] The earlier approaches were to learn the targeted CNN on both clean and adversarial samples as training set.…”