{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,8,2]],"date-time":"2025-08-02T16:55:57Z","timestamp":1754153757355,"version":"3.41.2"},"reference-count":0,"publisher":"ECMS","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,6,24]]},"abstract":"<jats:p>Machine learning (ML) techniques have gained\nwidespread adoption in medical image diagnosis.\nHowever, their susceptibility to adversarial attacks\nraises concerns regarding their reliability in clinical\napplications. This study investigates the robustness of two\nconvolutional neural network architectures, ResNet50\nand VGG16, against adversarial perturbations introduced\nvia the Fast Gradient Signed Method (FGSM) and\nDeepFool algorithms. An experimental evaluation was\nconducted using medical imaging data from the Lung\nImage Database Consortium (LIDC-IDRI), comprising\ncomputed tomography (CT) images annotated for lung\nlesions. Model performance was quantitatively assessed\nusing metrics derived from confusion matrices, including\naccuracy, precision, sensitivity, specificity, and F1-score.\nThe results demonstrate a significant vulnerability of\nboth the ResNet50 and VGG16 networks to adversarial\nmanipulations, resulting in considerable degradation of\nthe classification accuracy, particularly under higher\nperturbation magnitudes. To mitigate these vulnerabil-\nities, adversarial training employing FGSM-generated\nperturbations was implemented, notably enhancing\nmodel robustness and classification performance in\nadversarial settings. The findings confirm the efficacy\nof adversarial training as a defensive approach against\nadversarial attacks; however, further research into\nadvanced adversarial defense mechanisms and novel\nmodel architectures remains essential to ensure the\nsecure and reliable deployment of ML models in medical\ndiagnostics.<\/jats:p>","DOI":"10.7148\/2025-0255","type":"proceedings-article","created":{"date-parts":[[2025,7,23]],"date-time":"2025-07-23T12:36:52Z","timestamp":1753274212000},"page":"255-261","source":"Crossref","is-referenced-by-count":0,"title":["Vulnerabilities Of Machine Learning Algorithms To Adversarial Attacks In Medical Images"],"prefix":"10.7148","author":[{"given":"Karolina","family":"Krzton","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Joanna","family":"Kolodziej","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Adrian","family":"Widlak","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Mateusz","family":"Nawrocki","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jose","family":"Sigut","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"4144","published-online":{"date-parts":[[2025,6,24]]},"event":{"name":"39th ECMS International Conference on Modelling and Simulation"},"container-title":["ECMS 2025 Proceedings edited by Marco Scarpa, Salvatore Cavalieri, Salvatore Serrano, Fabrizio De Vita"],"original-title":[],"deposited":{"date-parts":[[2025,7,23]],"date-time":"2025-07-23T12:36:53Z","timestamp":1753274213000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.scs-europe.net\/dlib\/2025\/2025-0255.html"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,6,24]]},"references-count":0,"URL":"https:\/\/doi.org\/10.7148\/2025-0255","relation":{},"subject":[],"published":{"date-parts":[[2025,6,24]]}}}