Search

Recurrent neural networks (RNNs) have been applied to a broad range of applications, including natural language processing, drug discovery, and video recognition. Their vulnerability to input perturbation is also known. Aligning with a view from software defect detection, this article aims to develop a coverage-guided testing approach to systematically exploit the internal behavior of RNNs, with the expectation that such testing can detect defects with high possibility. Technically, the long short-term memory network (LSTM), a major class of RNNs, is thoroughly studied. A family of three test metrics are designed to quantify not only the values but also the temporal relations (including both stepwise and bounded-length) exhibited when LSTM processing inputs. A genetic algorithm is applied to efficiently generate test cases. The test metrics and test case generation algorithm are implemented into a tool testRNN, which is then evaluated on a set of LSTM benchmarks. Experiments confirm that testRNN has advantages over the state-of-the-art tool DeepStellar and attack-based defect detection methods, owing to its working with finer temporal semantics and the consideration of the naturalness of input perturbation. Furthermore, testRNN enables meaningful information to be collected and exhibited for users to understand the testing results, which is an important step toward interpretable neural network testing. [ABSTRACT FROM AUTHOR]

Symbolic execution is a powerful technique for automating software testing to detect many types of errors such as memory errors and assertion violations. However, it encounters the problem of path explosion, and by using only assertions, it still lacks the capability of going deep into checking the functional correctness of a path based on corresponding formal specifications. To address these problems, we propose a specification-based incremental testing method with symbolic execution, called SIT-SE, providing a much more rigorous way to automatically check the functional correctness of all the discovered program paths, by introducing theorems (instead of assertions) for path correctness and branch sequence coverage algorithm for guiding a moderate path exploration. Compared with Hoare logic for proving the correctness of an entire program, a theorem in the SIT-SE is made for verifying the correctness of a program path. The proposed method carefully treats the relationship between a path condition and the specification in a theorem to restrict the monotonous path exploration, whereas traditional concolic testing methods roughly use one test data to determine the path correctness by assertions during long path searching. We use a classic case to demonstrate how the method works and conduct an experiment to evaluate the performance of both the proposed method and the commonly used well-known concolic testing tool KLEE. The experimental results show that our method SIT-SE is effective and outperforms KLEE in detecting faulty paths based on specifications. [ABSTRACT FROM AUTHOR]

Spectrum-based fault localization (SBFL) techniques can automatically localize software faults. They employ the program spectrum, such as code coverage profile with test verdicts, to rank the program entities based on their code suspiciousness. In the past decades, researchers have proposed many approaches to optimize these techniques; however, the program structure, which can influence their performance, is not taken into consideration in developing and improving these techniques. In this article, we identify and analyze the effect of the program structure on the application of SBFL techniques. We observe that some specific program structures may introduce structure bias to code suspiciousness and negatively influence the output of SBFL techniques. To mitigate these effects and improve the performance of fault localization, we propose Delta4Ts, a structure-aware technique. Delta4Ts references debugging history to alleviate the impact of structure bias in the calculation of code suspiciousness. It reasons from the observable suspicious value towards the desired suspicious value and the impact of structure bias. To evaluate Delta4Ts under practical constraints, we conduct a controlled experiment using nine widely-studied SBFL formulae on 12 C programs and 6 Java programs. The experiment results show that Delta4Ts can significantly improve the accuracy of the studied SBFL formulae by an average of 34.8% on 12 C programs and 30.6% on 6 Java programs, and improve more on subject programs associated with more history versions or having larger code sizes. [ABSTRACT FROM AUTHOR]

A geographic information system (GIS) provides basic location-enabled services for many different applications related to navigation, education, and telecommunications. It is a foundation for analysis and visualization. Testing GIS is critical, but challenging due to the difficulty to assess the correctness of GIS outputs, which is called the test oracle problem of software testing. Metamorphic testing alleviates the problem by constructing metamorphic relations (MRs) among multiple inputs and outputs of the program under test. In this article, a semiautomated metamorphic testing (SAMT) method, based on the formal MR model and an improved adaptive random testing algorithm, was proposed to the GIS. To evaluate the performance of our approach, we conducted a case study on a superficial area calculation program, a typical component of GIS. Six kinds of MR construction methods were suggested for the GIS domain program testing. The experimental results show that SAMT can detect the mutations effectively that could solve the test oracle problem efficiently. More importantly, there is no need to manual participation in the testing process, except for the MR construction. [ABSTRACT FROM AUTHOR]

Adaptive random testing (ART) was developed as an enhanced version of random testing to increase the effectiveness of detecting failures in programs by spreading the test cases evenly over the input space. However, heavy computation may be incurred. In this paper, three enhanced algorithms for fixed-size-candidate-set ART (FSCS-ART) are proposed based on the $k$ -dimensional tree (KD-tree) structure. The first algorithm Naive-KDFC constructs a KD-tree by splitting the input space with respect to every dimension successively in a round-robin fashion. The second algorithm SemiBal-KDFC improves the balance of the KD-tree by prioritizing the splitting according to the spread in each dimension. In order to control the number of traversed nodes in backtracking, the third algorithm LimBal-KDFC introduces an upper bound for the nodes involved. Simulation and empirical studies have been conducted to investigate the efficiency and effectiveness of the three algorithms. The experimental results show that these algorithms significantly reduce the computation time of the original FSCS-ART for low dimensions and for the case of high dimensions with low failure rates. The efficiency of SemiBal-KDFC is better than that of Naive-KDFC when the dimension is no more than 8, but LimBal-KDFC is the most efficient of all three. Although the limited backtracking leads only to an approximate nearest neighbor in LimBal-KDFC, its failure-detection effectiveness is, in fact, better than FSCS-ART in high-dimensional input spaces and has no significant deterioration in low-dimensional spaces. [ABSTRACT FROM AUTHOR]

Random testing and partition testing are two major families of software testing techniques. They have been compared both theoretically and empirically in numerous studies for decades, and it has been widely acknowledged that they have their own advantages and disadvantages and that their innate characteristics are fairly complementary to each other. Some work has been conducted to develop advanced testing techniques through the integration of random testing and partition testing, attempting to preserve the advantages of both while minimizing their disadvantages. In this paper, we propose a new testing approach, adaptive partition testing, where test cases are randomly selected from some partition whose probability of being selected is adaptively adjusted along the testing process. We particularly develop two algorithms, Markov-chain based adaptive partition testing and reward-punishment based adaptive partition testing, to implement the proposed approach. The former algorithm makes use of Markov matrix to dynamically adjust the probability of a partition to be selected for conducting tests; while the latter is based on a reward and punishment mechanism. We conduct empirical studies to evaluate the performance of the proposed algorithms using ten faulty versions of three large-scale open source programs. Our experimental results show that, compared with two baseline techniques, namely random partition testing (RPT) and dynamic random testing (DRT), our algorithms deliver higher fault-detection effectiveness with lower test case selection overhead. It is demonstrated that the proposed adaptive partition testing is an effective testing approach, taking advantages of both random testing and partition testing. [ABSTRACT FROM AUTHOR]

Previous research in defect prediction has proposed approaches to determine which files require additional testing resources. However, practitioners typically create tests at a higher level of abstraction, which may span across many files. In this paper, we study software testing, especially test resource prioritization, from a different perspective. We use topic models to generate topics that provide a high-level view of a system, allowing developers to look at the test case coverage from a different angle. We propose measures of how well tested and defect prone a topic is, allowing us to discover which topics are well tested and which are defect prone. We conduct case studies on the histories of Mylyn, Eclipse, and NetBeans. We find that 34–78% of topics are shared between source code and test files, indicating that we can use topic models to study testing; well-tested topics are usually less defect prone, defect-prone topics are usually undertested; we can predict which topics are defect prone but not well tested with an average precision and recall of 75% and 77%, respectively; our approach complements traditional prediction-based approaches by saving testing and code inspection effort; and our approach is not particularly sensitive to the parameters that we use. [ABSTRACT FROM PUBLISHER]

Random testing is a fundamental testing technique that can be used to generate test cases for both hardware and software systems. Quasi-random testing was proposed as an enhancement to the cost-effectiveness of random testing: In addition to having similar computation overheads to random testing, it makes use of quasi-random sequences to generate low-discrepancy and low-dispersion test cases that help deliver high failure-detection effectiveness. Currently, few algorithms exist to generate quasi-random sequences, and these are mostly deterministic, rather than random. A previous study of quasi-random testing has examined two methods for randomizing quasi-random sequences to improve their applicability in testing. However, these randomization methods still have shortcomings—one method does not introduce much randomness to the test cases, while the other does not support incremental test case generation. In this paper, we present an innovative approach to incrementally randomizing quasi-random sequences. The test cases generated by this new approach show a high degree of randomness and evenness in distribution. We also conduct simulations and empirical studies to demonstrate the applicability and effectiveness of our approach in software testing. [ABSTRACT FROM PUBLISHER]

Testing is a key activity of software development and maintenance that determines the level of reliability. Traceability is the ability to describe and follow the life of software artifacts, and has been promoted as a means for supporting various activities, most importantly testing. Traceability information facilitates the testing and debugging of complex software by modeling the dependencies between code and tests. Actively supplementing traceability to testing enables rectifying defects more reliably and efficiently. Despite its importance, the development of test-to-code traceability has not been sufficiently addressed in the literature, and even worse there is currently no organized review of traceability studies in this field. In this work, we have investigated the main conferences, workshops, and journals of the requirements engineering, testing, and reliability, and identified those contributions that refer to traceability topics. From that starting point, we characterized and analyzed the chosen contributions against three research questions by utilizing a comparative framework including nine criteria. As a result, our study arrives to some interesting points, and outlines a number of potential research directions. This, in turn, can pave the way for facilitating and empowering traceability research in this domain to assist software engineers and testers in test management. [ABSTRACT FROM AUTHOR]

Background: Cross-company defect prediction (CCDP) is a field of study where an organization lacking enough local data can use data from other organizations for building defect predictors. To support CCDP, data must be shared. Such shared data must be privatized, but that privatization could severely damage the utility of the data. Aim: To enable effective defect prediction from shared data while preserving privacy. Method: We explore privatization algorithms that maintain class boundaries in a dataset. CLIFF is an instance pruner that deletes irrelevant examples. MORPH is a data mutator that moves the data a random distance, taking care not to cross class boundaries. CLIFF+MORPH are tested in a CCDP study among 10 defect datasets from the PROMISE data repository. Results: We find: 1) The CLIFFed+MORPHed algorithms provide more privacy than the state-of-the-art privacy algorithms; 2) in terms of utility measured by defect prediction, we find that CLIFF+MORPH performs significantly better. Conclusions: For the OO defect data studied here, data can be privatized and shared without a significant degradation in utility. To the best of our knowledge, this is the first published result where privatization does not compromise defect prediction. [ABSTRACT FROM AUTHOR]

Modern vehicles have evolved to support connected and self-driving capabilities. The concepts such as connected driving, cooperative driving, and intelligent transportation systems have resulted in an increase in the connectivity of vehicles and subsequently created new information security risks. The original vehicular ad-hoc network term is now emerged to a new term, Internet of Vehicles (IoV), which is a typical application of symmetry of Internet of Things (IoT). Vehicle manufacturers address some critical issues such as software bugs or security issues through remote updates, and this gives rise to concerns regarding the security of updated components. Moreover, aftermarket units such as those imposed by transportation authorities or insurance companies expose vehicles to high risk. Software testing aims to ensure that software products are reliable and behave as expected. Many commercial and open-source software products undergo formal certifications to increase users' confidence in their accuracy, reliability, and security. There are different techniques for software certification, including test-based certification. Testcase repositories are available to support software testing and certification, such as the Linux Test Project for Linux kernel testing. Previous studies performed various testing and experimental evaluation of different parts of modern vehicles to assess the security risks. Due to the lack of trusted testcase repositories and a common approach for testing, testing efforts are performed individually. In this paper, we propose a blockchain-based approach for a testcase repository to support test-based software and security testing and overcome the lack of trusted testcase repositories. The novel concept Proof-of-Validation to manage global state is proposed to manage updates to the repository. The initial work in this study considers the LTP test suite as a use case for the testcase repository. This research work is expected to contribute to the further development in including evidence generation for testing verification. [ABSTRACT FROM AUTHOR]

Random testing is a basic software testing technique that can be used to assess the software reliability as well as to detect software failures. Adaptive random testing has been proposed to enhance the failure-detection capability of random testing. Previous studies have shown that adaptive random testing can use fewer test cases than random testing to detect the first software failure. In this paper, we evaluate and compare the performance of adaptive random testing and random testing from another perspective, that of code coverage. As shown in various investigations, a higher code coverage not only brings a higher failure-detection capability, but also improves the effectiveness of software reliability estimation. We conduct a series of experiments based on two categories of code coverage criteria: structure-based coverage, and fault-based coverage. Adaptive random testing can achieve higher code coverage than random testing with the same number of test cases. Our experimental results imply that, in addition to having a better failure-detection capability than random testing, adaptive random testing also delivers a higher effectiveness in assessing software reliability, and a higher confidence in the reliability of the software under test even when no failure is detected. [ABSTRACT FROM PUBLISHER]

Test case prioritization is the process of ordering the execution of test cases to achieve a certain goal, such as increasing the rate of fault detection. Increasing the rate of fault detection can provide earlier feedback to system developers, improving fault fixing activity and, ultimately, software delivery. Many existing test case prioritization techniques consider that tests can be run in any order. However, due to functional dependencies that may exist between some test cases—that is, one test case must be executed before another—this is often not the case. In this paper, we present a family of test case prioritization techniques that use the dependency information from a test suite to prioritize that test suite. The nature of the techniques preserves the dependencies in the test ordering. The hypothesis of this work is that dependencies between tests are representative of interactions in the system under test, and executing complex interactions earlier is likely to increase the fault detection rate, compared to arbitrary test orderings. Empirical evaluations on six systems built toward industry use demonstrate that these techniques increase the rate of fault detection compared to the rates achieved by the untreated order, random orders, and test suites ordered using existing "coarse-grained” techniques based on function coverage. [ABSTRACT FROM AUTHOR]

The growing dependence of society on software systems places a high premium on their reliable operation. Moreover, the stringent reliability expectations imposed on these systems must be achieved despite their increasing size and complexity, and decreasing resources available for their development and maintenance. To mitigate these dual challenges, a systematic approach to guide the allocation of resources to the components of a software system is necessary. This paper presents an optimization framework which considers the contribution of each component to system reliability to determine the amount of effort to be allocated to each component, towards the ultimate objective of achieving the specified system reliability target with minimal effort. We assume that the contribution of a component to system reliability is governed by two factors: the system architecture, and the effort-reliability relationship of the component. This characterization is referred to as “architecture-based optimization” because it considers the system architecture explicitly in the effort allocation process. It is demonstrated that the architecture-based optimization framework outperforms other effort allocation strategies, including equal component weighting, and expert opinion. [ABSTRACT FROM AUTHOR]

This paper describes the authors' experience of teaching agile software development to students of computer science, software engineering, and other related disciplines, and comments on the implications of this and the lessons learned. It is based on the authors' eight years of experience in teaching agile software methodologies to various groups of students at different universities, in different cultural settings, and in a number of courses and seminars. It specifically discusses three different courses on agile software development, given in different teaching settings and at different levels, and briefly surveys variations to these courses given elsewhere. Based on the experience acquired, analyses and evaluations conducted, and current pedagogical trends at relevant university departments, the authors provide recommendations on how to overcome potential problems in teaching agile software development and make their adoption more effective. [ABSTRACT FROM AUTHOR]

In this paper, we propose two general frameworks for deriving several software reliability growth models based on a non-homogeneous Poisson process (NHPP) in the presence of imperfect debugging and error generation. The proposed models are initially formulated for the case when there is no differentiation between failure observation and fault removal testing processes, and then extended for the case when there is a clear differentiation between failure observation and fault removal testing processes. During the last three decades, many software reliability growth models (SRGM) have been developed to describe software failures as a random process, and can be used to evaluate development status during testing. With SRGM, software engineers can easily measure (or forecast) the software reliability (or quality), and plot software reliability growth charts. It is not easy to select the best model from a plethora of models available. There are few SRGM in the literature of software engineering that differentiates between failure observation and fault removal processes. In real software development environments, the number of failures observed need not be the same as the number of faults removed. Due to the complexity of software systems, and an incomplete understanding of software, the testing team may not be able to remove the fault perfectly on observation of a failure, and the original fault may remain, resulting in a phenomenon known as imperfect debugging, or get replaced by another fault causing error generation. In the case of imperfect debugging, the fault content of the software remains the same; while in the case of error generation, the fault content increases as the testing progresses. Removal of observed faults may result in the introduction of new faults. [ABSTRACT FROM AUTHOR]

Phillip G. Armour responds to "Differential Debugging" in the Tools of the Trade column September/October issue of IEEE Software to discuss the process of predicting defects. [ABSTRACT FROM PUBLISHER]

The problem of deciding whether an observed behavior is acceptable is the oracle problem. When testing from a finite state machine (FSM), it is easy to solve the oracle problem and so it has received relatively little attention for FSMs. However, if the system under test has physically distributed interfaces, called ports, then in distributed testing, we observe a local trace at each port and we compare the set of local traces with the set of allowed behaviors (global traces). This paper investigates the oracle problem for deterministic and nondeterministic FSMs and for two alternative definitions of conformance for distributed testing. We show that the oracle problem can be solved in polynomial time for the weaker notion of conformance (\sqsubseteq_w) but is NP-hard for the stronger notion of conformance (\sqsubseteq_s), even if the FSM is deterministic. However, when testing from a deterministic FSM with controllable input sequences, the oracle problem can be solved in polynomial time and similar results hold for nondeterministic FSMs. Thus, in some cases, the oracle problem can be efficiently solved when using \sqsubseteq_s and where this is not the case, we can use the decision procedure for \sqsubseteq_w as a sound approximation. [ABSTRACT FROM PUBLISHER]

This study is a quasi experiment comparing the software defect rates and implementation costs of two methods of software defect reduction: code inspection and test-driven development. We divided participants, consisting of junior and senior computer science students at a large Southwestern university, into four groups using a two-by-two, between-subjects, factorial design and asked them to complete the same programming assignment using either test-driven development, code inspection, both, or neither. We compared resulting defect counts and implementation costs across groups. We found that code inspection is more effective than test-driven development at reducing defects, but that code inspection is also more expensive. We also found that test-driven development was no more effective at reducing defects than traditional programming methods. [ABSTRACT FROM PUBLISHER]

Event-Driven Software (EDS) can change state based on incoming events; common examples are GUI and Web applications. These EDSs pose a challenge to testing because there are a large number of possible event sequences that users can invoke through a user interface. While valuable contributions have been made for testing these two subclasses of EDS, such efforts have been disjoint. This work provides the first single model that is generic enough to study GUI and Web applications together. In this paper, we use the model to define generic prioritization criteria that are applicable to both GUI and Web applications. Our ultimate goal is to evolve the model and use it to develop a unified theory of how all EDS should be tested. An empirical study reveals that the GUI and Web-based applications, when recast using the new model, show similar behavior. For example, a criterion that gives priority to all pairs of event interactions did well for GUI and Web applications; another criterion that gives priority to the smallest number of parameter value settings did poorly for both. These results reinforce our belief that these two subclasses of applications should be modeled and studied together. [ABSTRACT FROM AUTHOR]