Akram Kalaee

Soroush Hashemifar, Saeed Parsa, Akram Kalaee

Deep learning has revolutionized numerous fields, yet the reliability of Deep Neural Networks (DNNs) remains a concern due to their complexity and data dependency. Traditional software fault localization methods, such as Spectrum-based Fault Localization (SBFL), have been adapted for DNNs but often fall short in effectiveness. These methods typically overlook the propagation of faults through neural pathways, resulting in less precise fault detection. Research indicates that examining neural pathways, rather than individual neurons, is crucial because issues in one neuron can affect its entire pathway. By investigating these interconnected pathways, we can better identify and address problems arising from the collective activity of neurons. To address this limitation, we introduce the NP-SBFL method, which leverages Layer-wise Relevance Propagation (LRP) to identify essential faulty neural pathways. Our method explores multiple fault sources to accurately pinpoint faulty neurons by analyzing their interconnections. Additionally, our multi-stage gradient ascent (MGA) technique, an extension of gradient ascent (GA), enables sequential neuron activation to enhance fault detection. We evaluated NP-SBFL-MGA on the well-established MNIST and CIFAR-10 datasets, comparing it to other methods like DeepFault and NP-SBFL-GA, as well as three neuron measures: Tarantula, Ochiai, and Barinel. Our evaluation utilized all training and test samples (60,000 for MNIST and 50,000 for CIFAR-10) and revealed that NP-SBFL-MGA significantly outperformed the baselines in identifying suspicious pathways and generating adversarial inputs. Notably, Tarantula with NP-SBFL-MGA achieved a remarkable 96.75% fault detection rate compared to DeepFault's 89.90%. NP-SBFL-MGA highlights a strong correlation between critical path coverage and the number of failed tests in DNN fault localization.

Morteza Zakeri Nasrabadi, Saeed Parsa, Akram Kalaee

Appropriate test data is a crucial factor to reach success in dynamic software testing, e.g., fuzzing. Most of the real-world applications, however, accept complex structure inputs containing data surrounded by meta-data which is processed in several stages comprising of the parsing and rendering (execution). It makes the automatically generating efficient test data, to be non-trivial and laborious activity. The success of deep learning to cope in solving complex tasks especially in generative tasks has motivated us to exploit it in the context of complex test data generation. To do so, a neural language model (NLM) based on deep recurrent neural networks (RNNs) is used to learn the structure of complex input. Our approach generates new test data while distinguishes between data and meta-data that makes it possible to target both the parsing and rendering parts of software under test (SUT). Such test data can improve, input fuzzing. To assess the proposed approach, we developed a modular file format fuzzer, IUST-DeepFuzz. Our conducted experiments on the MuPDF, a lightweight and favorite portable document format (PDF) reader, reveal that IUST-DeepFuzz reaches high coverage of SUT in comparison with the state-of-the-art tools such as learn&fuzz, AFL, Augmented-AFL and random fuzzing. We also observed that the simpler deep learning models, the higher code coverage.