從COTS/Binary元件中解析條件參數以輔助隨機測試

Abstract

軟體品質透過軟體測試來達到驗證。軟體測試是想要找到程式缺陷而進行技巧性審查的過程。目前有許多的方式應用在軟體測試上，但是有些臭蟲是卻很難用傳統的測試方法來找到，假如他們不常發生在一般的情況下。為了改進軟體測試的涵蓋率，許多研究中提出自動化產生所有可能輸入資料的技術。 近期，較先進的做法是將具體和符號執行結合應用於隨機測試中。先簡化受測程式的原始碼成基本的指令型式，然後依據不同指令附加額外的程式碼以描述程式的行為，使用者提供具體的輸入值來執行轉換後的程式。在實際執行後，相關的條件參數會被儲存起來，然後再解析條件參數以產生測試的例子。但是這種方法不適合使用在沒有原始碼的程式上，特別是程式中使用到商業軟體或執行檔的部份。我們在ALERT（自動化隨機與邏輯推理運算測試）中提出一個解決之道，以接近原本程式語意的原則建立符號參數跟相依變數之間的關係。因此，可以保有原本符號的屬性，並且抵達更多的程式路徑。

Software quality is verified through software testing. Software testing is a process of technical investigation that is intended to reveal faults. There are many approaches to software testing, but bugs are difficult to find in conventional testing if they occur infrequently. In order to improve the test coverage, several techniques have been proposed to automatically generate all possible values of the inputs. Recently, the developed methods are combining concrete and symbolic execution for Random Testing. They first try to instrument the source code of a program under testing. The program is executed on some user-provided concrete input values. After the concrete run, symbolic constraints are stored and then generate concrete test cases by solving these symbolic constraints. Unfortunately, access to instrument source code is often infeasible especially for COTS/Binary component. We present a method in the framework, ALERT (Automatic Logic Evaluation for Random Testing), which approximates the program semantics and establish a connection between symbolic parameters and dependent variables. As a result, the symbolic property can be preserved and more paths will be reached.