Graphical User Interface Input Constraint Recommendation Based On CBA

Input validation is used to check whether the data received by the application meets the standards defined by the application.When input validation uses improper input constraints,applications are vulnerable to external malicious attacks.This brings a great negative impact on the security of software system,and also affects the use of users,even their personal life.Due to the limitation of developers' own thinking and the lack of security development awareness,they do not pay enough attention to input verification,so that improper input constraints are often added to input components in the process of software development,or even no input constraints are added.In order to avoid the generation of input verification vulnerabilities,the industry has spent a lot of energy on the research of input verification of software applications,and the current research mainly has two problems:(1)automated software testing,although it saves testing time and improves software testing efficiency.However,this does not solve the application with improper input constraints from the root.Developers still need to spend a lot of time and energy on improving and modifying the application.(2)As a part of software application input verification,input constraints play an important role in ensuring the security of software system and avoiding external attacks.However,due to their own development experience and thinking mode,software developers often add improper input constraints to the applied input verification function.Even experienced developers have difficulty remembering a large number of constraints.To solve the above problems,this paper focuses on the development phase of the project.When the developer implements the visual draft provided by the UI designer into the code,the input components in the GUI interface are identified and the appropriate input constraints are recommended.In this paper,the main research work and innovation are as follows:(1)Construction of the input constraint database.At present,there is no related research to build datasets for input constraints.Therefore,this paper extracts input constraints from developed front-end projects,at the same time,obtains the order of textview in the page form,the language of label text,and the number of components in the form.Then the label semantics of the textview is represented by the word embedding,and the Bag-of-word model represents the functional scene of the page,and the label text and the category of the page are obtained by clustering analysis.Finally,these features of textview are integrated to construct an input constraint database,which is the basis of subsequent input constraint recommendations.(2)For GUI text elements and non-text elements have different visual features,so that the accuracy of textview recognition is not high.A method for detecting and identifying text and non-text elements in GUI is proposed.When detecting non-text elements,this paper uses traditional computer vision technology to segment the GUI interface until only one GUI element is included in each block.Then the image classifier model is used to determine the category of GUI elements.The GUI text elements are obtained by scene text detection technology.Experimental results show that the accuracy of the proposed method is 52.3% higher than that of the target detection model based on depth learning,which can detect more GUI elements and more accurate bounding boxes.(3)Based on the method of input constraint library and GUI interface element recognition,the association relation between the conditional attributes in the input constraint library is explored by using the classification association rule algorithm,and the input constraint classifier is constructed by using the excavated classification association rule.and then the recognition method of GUI interface elements is integrated with the classifier to realize the recommendation of input constraints.In this paper,the trained input classifier is evaluated by 10-fold crossing validation,and its error rate is 3.2,which basically meets the requirements of the classification task.The experimental results of input constraint recommendation show that the overall recommendation accuracy of the input constraint recommendation module is more than 60%,which meets the expectations.