Treffer: Debloating Software Through Enhanced Static Analysis and Constraint Rules.

Introduction: Java applications often bloat, consuming more resources than necessary. Existing bytecode debloating techniques have several limitations, such as compromised correctness caused by incomplete code collection, which leads to false positives. The debloating process is resource‐intensive because it relies significantly on the test coverage to identify unnecessary code. This method is particularly problematic for large‐enterprise applications, where executing a single test case can take hours or days. In addition, most available debloating tools are standalone utilities that require manual configuration to be integrated with the project‐building process. Methods: In this study, we introduce an automated approach known as Trimming. Trimming achieves three main improvements: (1) it collects the necessary code using enhanced static analysis, improves the modeling of reflective calls, and infers instantiation objects. It also uses constraint rules to identify and process reference codes that contribute to bloating but cannot be removed without causing errors. (2) It is not based on test coverage. (3) It allows for direct debloating from the project build by implementing a Maven plugin that is configured within the POM.xml file of the bloated project. Results: Our evaluation results show that when Java reflection analysis or constraint rules in TRIMMING were disabled, the effectiveness of debloating analysis dropped to 88.25% with a 95% confidence interval (CI) of [85.3%, 91.2%], and 35.0% with a 95% CI of [30.5%, 39.5%], respectively. Moreover, inferring instantiation objects reduced the program size by 4.9%. Trimming achieved a bytecode reduction rate of 24.1%, with all the debloated projects successfully compiling and passing test suites. [ABSTRACT FROM AUTHOR]

Copyright of Software: Practice & Experience is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)