Convolutional neural networks (CNNs) have been enormously successful in a variety of image recognition tasks. Robustness is an important metric to evaluate the quality of CNNs. However, recent research shows that CNNs are particularly vulnerable to adversarial attacks. This paper proposes an adversarial defense method to increase the robustness of CNNs, namely, SCADefender. The proposed method trains a reformer on adversarial examples and the training set of a target classifier. The architecture of the reformer is stacked convolutional autoencoder. The adversarial examples are generated by using various adversarial attacks such as untargeted FGSM, untargeted CW [Formula: see text] and untargeted BIS. Given an input image, the trained reformer could remove the adversarial perturbations with a low computational cost. To demonstrate the effectiveness, the proposed method is compared with PuVAE, MagNet, and adversarial training on three well-known datasets including MNIST, Fashion-MNIST, and CIFAR-10. In terms of the average detection rate, the proposed method outperforms other methods. While the proposed method achieves an average detection rate of 97.78% for MNIST, 90.43% for Fashion-MNIST, and 80.64% for CIFAR-10, the comparable methods achieve only 23.69- 86.18% for MNIST, 63.90-79.70% for Fashion-MNIST, and 25.55-77.36% for CIFAR-10.

This paper improves the breadth-first search strategy in directed automated random testing (DART) to generate a fewer number of test data while gaining higher branch coverage, namely Static DART or SDART for short. In addition, the paper extends the test data compilation mechanism in DART, which currently only supports the projects written in C, to generate test data for C++ projects. The main idea of SDART is when it is less likely to increase code coverage with the current path selection strategies, the static test data generation will be applied with the expectation that more branches are covered earlier. Furthermore, in order to extend the test data compilation of DART for C++ context, the paper suggests a general test driver technique for C++ which supports various types of parameters including basic types, arrays, pointers, and derived types. Currently, an experimental tool has been implemented based on the proposal in order to demonstrate its efficacy in practice. The results have shown that SDART achieves higher branch coverage with a fewer number of test data in comparison with that of DART in practice.

This paper proposes a framework for verifying component-based software in the context of component evolution. This framework includes two stages: modular conformance testing for updating inaccurate model of the evolved component and modular verification for evolving component-based software. When a component is evolved after adapting some refinements, the proposed framework only focuses on this component and its model in order to update the model and recheck the whole evolved system. The framework also reuses the previous verification results and the previous models of the evolved component to reduce several steps of the model update and verification processes.