Automated Strong Mutation Testing of XACML Policies
Automated Strong Mutation Testing of XACML Policies
复制标题
DOI：
10.1145/3381991.3395599
复制
发表时间：
2020-05
期刊：
影响因子：
--
通讯作者：
Dianxiang Xu;Roshan Shrestha;Ning Shen
中科院分区：
文献类型：
--
作者：
Dianxiang Xu;Roshan Shrestha;Ning Shen
虽然现有的测试XACML策略的方法具有不同程度的有效性，但它们都不能揭示大多数策略错误。未公开的故障可能导致未经授权的访问和拒绝服务。本文提出了一种XACML策略的强突变测试方法，该方法可以自动生成来自给定策略的突变体的测试。这样的突变体表示可能出现在策略中的目标故障。在这种方法中，我们首先组成的强突变约束，捕捉每个突变体之间的语义差异，其原始政策。然后，我们使用约束求解器来导出访问请求（即，test）。由策略的所有突变体生成的测试集可以获得完美的突变分数，从而发现所有假设的错误或证明它们不存在。在基于变异的方法的基础上，本文进一步探讨了最优的测试套件，达到一个完美的变异分数没有重复测试。为了评估所提出的方法，我们的实验包括所有的主题政策在相关文献中，并使用了一些新的政策。结果表明：（1）生成基于变异的测试集以获得完美的变异分数是可扩展的，（2）由于昂贵的重复测试的去除，生成最优测试集可能是不切实际的，（3）与现有研究的结果不同，基于修改条件/决策覆盖的方法，目前最有效的一种，对于几个策略具有低变异分数。
While the existing methods for testing XACML policies have varying levels of effectiveness, none of them can reveal the majority of policy faults. The undisclosed faults may lead to unauthorized access and denial of service. This paper presents an approach to strong mutation testing of XACML policies that automatically generates tests from the mutants of a given policy. Such mutants represent the targeted faults that may appear in the policy. In this approach, we first compose the strong mutation constraints that capture the semantic difference between each mutant and its original policy. Then, we use a constraint solver to derive an access request (i.e., test). The test suite generated from all the mutants of a policy can achieve a perfect mutation score, thus uncover all hypothesized faults or demonstrate their absence. Based on the mutation-based approach, this paper further explores optimal test suite that achieves a perfect mutation score without duplicate tests. To evaluate the proposed approach, our experiments have included all the subject policies in the relevant literature and used a number of new policies. The results demonstrate that: (1) it is scalable to generate a mutation-based test suite to achieve a perfect mutation score, (2) it can be impractical to generate the optimal test suite due to the expensive removal of duplicate tests, (3) different from the results of the existing study, the modified-condition/decision coverage-based method, currently the most effective one, has low mutation scores for several policies.