CINXE.COM

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"> <html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8" /> <title>ComSIS | Computer Science and Information Systems</title> <link rel="stylesheet" type="text/css" href="res/style1.css" /> </head> <body> <script type="text/javascript" src="res/wz_tooltip.js"></script> <script type="text/javascript" src="res/slide.js"></script> <div id="all"> <div id="header"> <h1>Computer Science and Information Systems</h1> </div>  <div id="main"> <div id="sidebar"> <p>About the journal</p> <ul> <li><a href="index.php">Home page</a></li> <li><a href="contact.php">Contact information</a></li> <li><a href="aims.php">Aims and scope</a></li> <li><a href="indexing.php">Indexing information</a></li> <li><a href="policies.php">Editorial policies</a></li> <li><a href="consortium.php">ComSIS consortium</a></li> <li><a href="boards.php">Journal boards</a></li> <li><a href="managing.php">Managing board</a></li> </ul> <p>For authors</p> <ul> <li><a href="information.php">Information for contributors</a></li> <li><a href="http://ojs.pmf.uns.ac.rs/index.php/comsis">Paper submission</a></li> <li><a href="submission.php">Article submission through OJS</a></li> <li><a href="copyright.php">Copyright transfer form</a></li> <li><a href="download.php">Download section</a></li> </ul> <p>For readers</p> <ul> <li><a href="archive.php?show=lstnew">Forthcoming articles</a></li> <li><a href="archive.php?show=vol2104">Current issue</a></li> <li><a href="archive.php">Archive</a></li> </ul> <p>For reviewers</p> <ul> <li><a href="http://ojs.pmf.uns.ac.rs/index.php/comsis">View and review submissions</a></li> </ul> <p>News</p> <ul> <li><a href="https://www.facebook.com/ComSISJournal/"> <img src="res/fb.png" alt="FB"/> Journal's Facebook page</a></li> <li><a href="cfp.php">Calls for special issues</a></li> <li><a href="notification.php">New issue notification</a></li> </ul> </div>  <div id="content">  <h1 class="title">Semantic Feature-Based Test Selection for Deep Neural Networks: A Frequency Domain Perspective</h1><p class="authors">Zhouxian Jiang<sup>1</sup>, Honghui Li<sup>1</sup>, Xuetao Tian<sup>2</sup> and Rui Wang<sup>1</sup></p><ol><li>School of Computing and Information Technology, Beijing Jiaotong University, Beijing, China</li><li>Faculty of Psychology, Beijing Normal University, Beijing, China</li></ol><h3>Abstract</h3><p>While deep neural networks (DNNs) have great potential for applications in security and safety-critical domains, their limited robustness to adversarial samples and out-of-distribution (OOD) samples raise significant concerns. In the software engineering community, significant efforts have been devoted to devising testing techniques that verify the robustness of DNNs. This paper investigates semantic feature-based test selection for DNNs from a frequency domain perspective and propose a novel method called SaFeTS. Specifically, we leverage saliency detection techniques, such as Fourier Phase Transform to extract semantic features from test cases. These features are then clustered to select diverse test cases to evaluate the robustness of DNNs and model retraining. Experiments on CIFAR-10 and SVHN datasets demonstrate that SaFeTS exposes more varied model errors compared to baseline methods. Further, retraining with SaFeTS-selected samples significantly improves adversarial and out-of-distribution robustness over state-of-the-art test selection methods.</p><h3>Key words</h3><p>DNN testing, test selection, semantic feature, frequency domain, robustness</p><h3>Digital Object Identifier (DOI)</h3><p><a href="https://doi.org/10.2298/CSIS230907045J">https://doi.org/10.2298/CSIS230907045J</a></p><h3>Publication information</h3><p><a href="/archive.php?show=vol2104">Volume 21, Issue 4 (September 2024)</a><br/>Year of Publication: 2024<br/>ISSN: 2406-1018 (Online)<br/>Publisher: ComSIS Consortium</p><h3>Full text</h3><p><a class="download" href="pdf.php?id=16483"><img class="left" src="res/pdf.png" alt="Download"/>Available in PDF<br/><em>Portable Document Format</em></a></p><h3>How to cite</h3><p>Jiang, Z., Li, H., Tian, X., Wang, R.: Semantic Feature-Based Test Selection for Deep Neural Networks: A Frequency Domain Perspective. Computer Science and Information Systems, Vol. 21, No. 4, 1499–1522. (2024), https://doi.org/10.2298/CSIS230907045J</p>  </div>  </div>  <div id="footer_top"> </div> <div id="footer"> <div class="left">Faculty of Sciences, Trg Dositeja Obradovića 3, 21000 Novi Sad, Serbia, <a href="mailto:comsis@uns.ac.rs">comsis@uns.ac.rs</a></div> <div class="left">Published by ComSIS Consortium under<br/><a rel="license" href="http://creativecommons.org/licenses/by-nc-nd/4.0/">Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License<br><img alt="Creative Commons License" style="border-width:0" src="https://i.creativecommons.org/l/by-nc-nd/4.0/88x31.png"/></a></div> <div class="clearer"> </div> </div>  </div>  </body> </html>