<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.3 20210610//EN" "JATS-journalpublishing1-3.dtd">
<article article-type="research-article" dtd-version="1.3" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xml:lang="ru"><front><journal-meta><journal-id journal-id-type="publisher-id">izvestswsu</journal-id><journal-title-group><journal-title xml:lang="ru">Известия Юго-Западного государственного университета</journal-title><trans-title-group xml:lang="en"><trans-title>Proceedings of the Southwest State University</trans-title></trans-title-group></journal-title-group><issn pub-type="ppub">2223-1560</issn><issn pub-type="epub">2686-6757</issn><publisher><publisher-name>ЮЗГУ</publisher-name></publisher></journal-meta><article-meta><article-id pub-id-type="doi">10.21869/2223-1560-2025-29-4-53-69</article-id><article-id custom-type="elpub" pub-id-type="custom">izvestswsu-1516</article-id><article-categories><subj-group subj-group-type="heading"><subject>Research Article</subject></subj-group><subj-group subj-group-type="section-heading" xml:lang="ru"><subject>ИНФОРМАТИКА, ВЫЧИСЛИТЕЛЬНАЯ ТЕХНИКА И УПРАВЛЕНИЕ</subject></subj-group><subj-group subj-group-type="section-heading" xml:lang="en"><subject>COMPUTER SCIENCE, COMPUTER ENGINEERING AND CONTROL</subject></subj-group></article-categories><title-group><article-title>Гибридный двухуровневый метод автоматического выявления подмены лица оператора на изображении</article-title><trans-title-group xml:lang="en"><trans-title>Hybrid two-level method for automatic detection of face substitution in an image</trans-title></trans-title-group></title-group><contrib-group><contrib contrib-type="author" corresp="yes"><name-alternatives><name name-style="eastern" xml:lang="ru"><surname>Халеев</surname><given-names>М. Д.</given-names></name><name name-style="western" xml:lang="en"><surname>Haleev</surname><given-names>M. D.</given-names></name></name-alternatives><bio xml:lang="ru"><p>Халеев Михаил Дмитриевич, кандидат технических наук, младший научный сотрудник</p><p>ул. Корпусная, д. 18, г. Санкт-Петербург 199178</p><p> </p></bio><bio xml:lang="en"><p>Mikhail D. Haleev, Cand. of Sci. (Engineering), Junior Research Fellow</p><p>18, Korpusnaya str., St. Petersburg 199178</p></bio><email xlink:type="simple">Haleev.M@iias.spb.su</email><xref ref-type="aff" rid="aff-1"/></contrib></contrib-group><aff-alternatives id="aff-1"><aff xml:lang="ru"><institution>Санкт-Петербургский Федеральный исследовательский центр Российской академии наук</institution></aff><aff xml:lang="en"><institution>St. Petersburg Federal Research Center of the Russian Academy of Sciences</institution></aff></aff-alternatives><pub-date pub-type="collection"><year>2025</year></pub-date><pub-date pub-type="epub"><day>08</day><month>01</month><year>2026</year></pub-date><volume>29</volume><issue>4</issue><fpage>53</fpage><lpage>69</lpage><permissions><copyright-statement>Copyright &amp;#x00A9; Халеев М.Д., 2026</copyright-statement><copyright-year>2026</copyright-year><copyright-holder xml:lang="ru">Халеев М.Д.</copyright-holder><copyright-holder xml:lang="en">Haleev M.D.</copyright-holder><license xml:lang="ru" license-type="creative-commons-attribution" xlink:href="https://creativecommons.org/licenses/by/4.0/" xlink:type="simple"><license-p>Данная работа распространяется под лицензией Creative Commons Attribution 4.0.</license-p></license><license xml:lang="en" license-type="creative-commons-attribution" xlink:href="https://creativecommons.org/licenses/by/4.0/" xlink:type="simple"><license-p>This work is licensed under a Creative Commons Attribution 4.0 License.</license-p></license></permissions><self-uri xlink:href="https://izvestswsu.elpub.ru/jour/article/view/1516">https://izvestswsu.elpub.ru/jour/article/view/1516</self-uri><abstract><sec><title>Цель исследования</title><p>Цель исследования: Разработка гибридного двухуровневого метода для повышения как точности, так и устойчивости выявления подмены лица оператора на изображениях, что является актуальной задачей в условиях постоянного роста и усложнения угроз со стороны дипфейк-технологий. </p></sec><sec><title>Методы</title><p>Методы. Предложена архитектура, объединяющая сверточную нейронную сеть EfficientNet для извлечения глубоких паттернов и ансамбль из четырех классификаторов. Эти классификаторы целенаправленно анализируют специфические группы признаков: экспертные, текстурные, статистические и основанные на координатах лицевых ориентиров, что позволяет выявлять конкретные артефакты синтеза. Для обучения и тестирования был сформирован обширный и репрезентативный комплексный набор данных объемом 34 000 изображений, включающий как сгенерированные дипфейки, так и публичные датасеты. </p></sec><sec><title>Результаты</title><p>Результаты. Экспериментально подтверждена высокая эффективность предложенного метода: точность составила 0,921, а F1-мера – 0,914. Эти показатели значительно превосходят результаты любой из моделей, использованных по отдельности, что доказывает ярко выраженный и практически значимый синергетический эффект от их объединения. </p></sec><sec><title>Заключение</title><p>Заключение. Работа демонстрирует, что синергия глубокого обучения и классических признаковых моделей позволяет создать действительно более надежный и точный детектор. Предложенный метод повышает общую точность и увеличивает надежность системы, эффективно компенсируя индивидуальные слабости отдельных классификаторов. Это подтверждает гипотезу о том, что сочетание способности нейросети извлекать сложные, неявные паттерны и способности признаковых моделей анализировать конкретные, заранее известные специфические артефакты (например, геометрические искажения) ведет к созданию более мощного и устойчивого детектора. </p></sec></abstract><trans-abstract xml:lang="en"><sec><title>Purpose of research</title><p>Purpose of research. The development of a hybrid, two-level method to enhance both the accuracy and robustness of detecting operator face spoofing in images, which is a pressing issue given the constant growth and sophistication of threats from deepfake technologies. </p></sec><sec><title>Methods</title><p>Methods. A novel architecture is proposed, combining the EfficientNet convolutional neural network for deep pattern extraction with an ensemble of four classifiers. These classifiers specifically analyze distinct feature groups: expertbased, textural, statistical, and those based on facial landmark coordinates, enabling the detection of specific synthesis artifacts. For training and testing, an extensive and representative dataset of 34,000 images was compiled, including deepfakes generated by several modern tools as well as public datasets. </p></sec><sec><title>Results</title><p>Results. The high efficacy of the proposed method was experimentally confirmed: accuracy reached 0.921 and the F1-score was 0.914. These metrics significantly surpass the performance of any of the individual models used separately, demonstrating a pronounced and practically significant synergistic effect from their combination. </p></sec><sec><title>Conclusion</title><p>Conclusion. This work demonstrates that the synergy between deep learning and classical feature-based models allows for the creation of a genuinely more reliable and precise detector. The proposed method improves overall accuracy and enhances system robustness by effectively compensating for the individual weaknesses of separate classifiers. This validates the hypothesis that combining a neural network's ability to extract complex, implicit patterns with feature-based models' capacity to analyze specific, predefined artifacts (such as geometric distortions) leads to a more powerful and resilient detector.</p></sec></trans-abstract><kwd-group xml:lang="ru"><kwd>подмена лиц</kwd><kwd>компьютерное зрение</kwd><kwd>искусственный интеллект</kwd><kwd>глубокое обучение</kwd><kwd>дипфейк</kwd><kwd>информационная безопасность</kwd></kwd-group><kwd-group xml:lang="en"><kwd>face swapping</kwd><kwd>computer vision</kwd><kwd>artificial intelligence</kwd><kwd>deep learning</kwd><kwd>deepfake</kwd><kwd>information security</kwd></kwd-group><funding-group><funding-statement xml:lang="ru">Исследования выполнены в рамках бюджетной темы FFZF-2025-0003.</funding-statement><funding-statement xml:lang="en">Research was supported by Russian State Research FFZF-2025-0003.</funding-statement></funding-group></article-meta></front><back><ref-list><title>References</title><ref id="cit1"><label>1</label><citation-alternatives><mixed-citation xml:lang="ru">Халеев М.Д. Интеллектуальный метод автоматического выявления подмены лица на изображении // Системы анализа и обработки данных. 2025. Т. 97, №1. С. 105-120.</mixed-citation><mixed-citation xml:lang="en">Haleev M. D. Intelligent method for automatic detection of face substitution in an image. Sistemy analiza i obrabotki dannykh = Analysis and Data Processing Systems. 2025; 97(1): 105-120. (In Russ.).</mixed-citation></citation-alternatives></ref><ref id="cit2"><label>2</label><citation-alternatives><mixed-citation xml:lang="ru">Tolosana R., Vera-Rodriguez R., Fierrez J., Morales A., Ortega-Garcia J. Deepfakes and beyond: A Survey of face manipulation and fake detection // Information Fusion. 2020. Vol. 64. P. 131–148. https://doi:10.1016/J.INFFUS.2020.06.014.</mixed-citation><mixed-citation xml:lang="en">Tolosana R., Vera-Rodriguez R., Fierrez J., Morales A.,  Ortega-Garcia J. Deepfakes and beyond: A Survey of face manipulation and fake detection. Information Fusion. 2020; 64: 131–148.  https://doi:10.1016/J.INFFUS.2020.06.014.</mixed-citation></citation-alternatives></ref><ref id="cit3"><label>3</label><citation-alternatives><mixed-citation xml:lang="ru">Nawaz M., Javed A., Irtaza A. A deep learning model for FaceSwap and facereenactment deepfakes detection // Applied Soft Computing. 2024. Vol. 162. P. 111854. https://doi:10.1016/J.ASOC.2024.111854.</mixed-citation><mixed-citation xml:lang="en">Nawaz M., Javed A., A. Irtaza A deep learning model for FaceSwap and facereenactment deepfakes detection. Appl Soft Comput. 2024; 162: 111854. https://doi:10.1016/J.ASOC.2024.111854.</mixed-citation></citation-alternatives></ref><ref id="cit4"><label>4</label><citation-alternatives><mixed-citation xml:lang="ru">Ding X. и др. Swapped face detection using deep learning and subjective assessment // EURASIP Journal on Information Security. 2020. Vol. 2020, № 1. https://doi:10.1186/S13635-020-00109-8.</mixed-citation><mixed-citation xml:lang="en">Ding X., Raziei Z., Larson E. C., E Olinick. V., Krueger P., Hahsler M. Swapped face detection using deep learning and subjective assessment. EURASIP J Inf Secur. 2020; 2020(1). https://doi:10.1186/S13635-020-00109-8.</mixed-citation></citation-alternatives></ref><ref id="cit5"><label>5</label><citation-alternatives><mixed-citation xml:lang="ru">Essa E. Feature fusion Vision Transformers using MLP-Mixer for enhanced deepfake detection // Neurocomputing. 2024. Vol. 598. P. 128128. https://doi:10.1016/J.NEUCOM.2024.128128</mixed-citation><mixed-citation xml:lang="en">Essa E. Feature fusion Vision Transformers using MLP-Mixer for enhanced deepfake detection. Neurocomputing. 2024; 598: 128128. https://doi:10.1016/J.NEUCOM.2024.128128.</mixed-citation></citation-alternatives></ref><ref id="cit6"><label>6</label><citation-alternatives><mixed-citation xml:lang="ru">Salman M., et al. AWARE-NET: Adaptive Weighted Averaging for Robust Ensemble Network in Deepfake Detection // Computer Vision and Pattern Recognition. 2025.</mixed-citation><mixed-citation xml:lang="en">Muhammad Salman, Iqra Tariq, Mishal Zulfiqar, Muqadas Jalal, Sami Aujla, Sumbal Fatima, AWARE-NET: Adaptive Weighted Averaging for Robust Ensemble Network in Deepfake Detection. Computer Vision and Pattern Recognition. 2025.</mixed-citation></citation-alternatives></ref><ref id="cit7"><label>7</label><citation-alternatives><mixed-citation xml:lang="ru">Kingra S., Aggarwal N., Kaur N. SFormer: An end-to-end spatio-temporal transformer architecture for deepfake detection // Forensic Science International: Digital Investigation. 2024. Vol. 51. P. 301817. https://doi:10.1016/J.FSIDI.2024.301817</mixed-citation><mixed-citation xml:lang="en">Kingra S., N Aggarwal., Kaur N. SFormer: An end-to-end spatio-temporal transformer architecture for deepfake detection. Forensic Science International: Digital Investigation. 2024; 51: 301817. https://doi:10.1016/J.FSIDI.2024.301817.</mixed-citation></citation-alternatives></ref><ref id="cit8"><label>8</label><citation-alternatives><mixed-citation xml:lang="ru">Khalid F., Javed A., ul ain Q., Ilyas H., Irtaza A. DFGNN: An interpretable and generalized graph neural network for deepfakes detection // Expert Systems with Applications. 2023. Vol. 222. P. 119843. https://doi:10.1016/J.ESWA.2023.119843</mixed-citation><mixed-citation xml:lang="en">Khalid F., Javed A., ul ain Q., Ilyas H., Irtaza A. DFGNN: An interpretable and generalized graph neural network for deepfakes detection. Expert Syst Appl. 2023; 222: 119843.  https://doi:10.1016/J.ESWA.2023.119843.</mixed-citation></citation-alternatives></ref><ref id="cit9"><label>9</label><citation-alternatives><mixed-citation xml:lang="ru">Sun K., et al. DiffusionFake: Enhancing Generalization in Deepfake Detection via Guided Stable Diffusion // 38th Conference on Neural Information Processing Systems (NeurIPS 2024). 2024.</mixed-citation><mixed-citation xml:lang="en">Ke Sun, Shen Chen, Taiping Yao, Hong Liu, Xiaoshuai Sun, Shouhong Ding, Rongrong Ji, DiffusionFake: Enhancing Generalization in Deepfake Detection via Guided Stable Diffusion. 38th Conference on Neural Information Processing Systems (NeurIPS 2024). 2024.</mixed-citation></citation-alternatives></ref><ref id="cit10"><label>10</label><citation-alternatives><mixed-citation xml:lang="ru">Smeu S., Oneata E., Oneata D. DeCLIP: Decoding CLIP Representations for Deepfake Localization // Proceedings of the Winter Conference on Applications of Computer Vision (WACV). 2025. С. 149-159.</mixed-citation><mixed-citation xml:lang="en">Stefan Smeu, Elisabeta Oneata, Dan Oneata, DeCLIP: Decoding CLIP Representations for Deepfake Localization. Proceedings of the Winter Conference on Applications of Computer Vision (WACV). 2025. P. 149-159.</mixed-citation></citation-alternatives></ref><ref id="cit11"><label>11</label><citation-alternatives><mixed-citation xml:lang="ru">Tian J., et al. Real Appearance Modeling for More General Deepfake Detection // ECCV. 2025.</mixed-citation><mixed-citation xml:lang="en">Jiahe Tian, Cai Yu, Xi Wang, Peng Chen, Zihao Xiao, Jiao Dai, Jizhong Han, and Yesheng Chai, Real Appearance Modeling for More General Deepfake Detection. ECCV, 2025.</mixed-citation></citation-alternatives></ref><ref id="cit12"><label>12</label><citation-alternatives><mixed-citation xml:lang="ru">Yan B., Li C. T., Lu X. JRC: Deepfake detection via joint reconstruction and classification // Neurocomputing. 2024. Vol. 598. P. 127862. https://doi:10.1016/J.NEUCOM.2024.127862</mixed-citation><mixed-citation xml:lang="en">Yan B., Li C. T., Lu X. JRC: Deepfake detection via joint reconstruction and classification, Neurocomputing. 2024; 598: 127862. https://doi:10.1016/J.NEUCOM.2024.127862.</mixed-citation></citation-alternatives></ref><ref id="cit13"><label>13</label><citation-alternatives><mixed-citation xml:lang="ru">Li H., et al. FreqBlender: Enhancing DeepFake Detection by Blending Frequency Knowledge // 38th Conference on Neural Information Processing Systems (NeurIPS 2024). 2024.</mixed-citation><mixed-citation xml:lang="en">Hanzhe Li, Jiaran Zhou, Yuezun Li, Baoyuan Wu, Bin Li, Junyu Dong, FreqBlender: Enhancing DeepFake Detection by Blending Frequency Knowledge. 38th Conference on Neural Information Processing Systems (NeurIPS 2024). 2024.</mixed-citation></citation-alternatives></ref><ref id="cit14"><label>14</label><citation-alternatives><mixed-citation xml:lang="ru">Kashiani H., Talemi N. A., Afghah F. FreqDebias: Towards Generalizable Deepfake Detection via Consistency-Driven Frequency Debiasing // CVPR. 2025.</mixed-citation><mixed-citation xml:lang="en">Hossein Kashiani, Niloufar Alipour Talemi, Fatemeh Afghah, FreqDebias: Towards Generalizable Deepfake Detection via Consistency-Driven Frequency Debiasing. CVPR, 2025.</mixed-citation></citation-alternatives></ref><ref id="cit15"><label>15</label><citation-alternatives><mixed-citation xml:lang="ru">Zou M., et al. Semantic Contextualization of Face Forgery: A New Definition, Dataset, and Detection Method // Computer Vision and Pattern Recognition. 2025.</mixed-citation><mixed-citation xml:lang="en">Mian Zou, Baosheng Yu, Yibing Zhan, Siwei Lyu, and Kede Ma, Semantic Contextualization of Face Forgery: A New Definition, Dataset, and Detection Method. Computer Vision and Pattern Recognition, 2025.</mixed-citation></citation-alternatives></ref><ref id="cit16"><label>16</label><citation-alternatives><mixed-citation xml:lang="ru">Chew C. J., et al. Preserving manipulated and synthetic Deepfake detection through face texture naturalness // Journal of Information Security and Applications. 2024. Vol. 83. P. 103798. https://doi:10.1016/J.JISA.2024.103798.</mixed-citation><mixed-citation xml:lang="en">Chew C. J., Lin Y. C., Chen Y. C., Fan Y. Y., Lee J. S. Preserving manipulated and synthetic Deepfake detection through face texture naturalness. Journal of Information Security and Applications. 2024; 83: 103798. https://doi:10.1016/J.JISA.2024.103798.</mixed-citation></citation-alternatives></ref><ref id="cit17"><label>17</label><citation-alternatives><mixed-citation xml:lang="ru">Gao J., at al. Texture and artifact decomposition for improving generalization in deep-learning-based deepfake detection // Engineering Applications of Artificial Intelligence. 2024. Vol. 133. P. 108450. https://doi:10.1016/J.ENGAPPAI.2024.108450</mixed-citation><mixed-citation xml:lang="en">Gao J., et al. Texture and artifact decomposition for improving generalization in deep-learning-based deepfake detection. Eng Appl Artif Intell. 2024; 133: 108450. https://doi:10.1016/J.ENGAPPAI.2024.108450.</mixed-citation></citation-alternatives></ref><ref id="cit18"><label>18</label><citation-alternatives><mixed-citation xml:lang="ru">He Q., Peng C., Liu D., Wang N., Gao X. GazeForensics: DeepFake detection via gaze-guided spatial inconsistency learning // Neural Networks. 2024. Vol. 180. P. 106636. https://doi:10.1016/J.NEUNET.2024.106636</mixed-citation><mixed-citation xml:lang="en">He Q., Peng C., Liu D., Wang N., Gao X. GazeForensics: DeepFake detection via gaze-guided spatial inconsistency learning. Neural Networks. 2024; 180: 106636. https://doi:10.1016/J.NEUNET.2024.106636.</mixed-citation></citation-alternatives></ref><ref id="cit19"><label>19</label><citation-alternatives><mixed-citation xml:lang="ru">Li Y., at al. Texture Shape and Order Matter: A New Transformer Design for Sequential DeepFake Detection // Proceedings of the Winter Conference on Applications of Computer Vision (WACV). 2025. P. 202-211.</mixed-citation><mixed-citation xml:lang="en">Yunfei Li, Yuezun Li, Xin Wang, Baoyuan Wu, Jiaran Zhou, Junyu Dong, Texture Shape and Order Matter: A New Transformer Design for Sequential DeepFake Detection. Proceedings of the Winter Conference on Applications of Computer Vision (WACV), 2025.  P. 202-211.</mixed-citation></citation-alternatives></ref><ref id="cit20"><label>20</label><citation-alternatives><mixed-citation xml:lang="ru">Wang Y., Huang H. Audio–visual deepfake detection using articulatory representation learning // Computer Vision and Image Understanding. 2024. Vol. 248. P. 104133. https://doi:10.1016/J.CVIU.2024.104133</mixed-citation><mixed-citation xml:lang="en">Wang Y.,  H. Huang Audio–visual deepfake detection using articulatory representation learning. Computer Vision and Image Understanding. 2024; 248: 104133. https://doi:10.1016/J.CVIU.2024.104133.</mixed-citation></citation-alternatives></ref><ref id="cit21"><label>21</label><citation-alternatives><mixed-citation xml:lang="ru">Wang T., Cheng H., Zhang X., Wang Y. NullSwap: Proactive Identity Cloaking Against Deepfake Face Swapping // CVPR. 2025.</mixed-citation><mixed-citation xml:lang="en">Tianyi Wang, Harry Cheng, Xiao Zhang, Yinglong Wang, NullSwap: Proactive Identity Cloaking Against Deepfake Face Swapping, CVPR, 2025.</mixed-citation></citation-alternatives></ref><ref id="cit22"><label>22</label><citation-alternatives><mixed-citation xml:lang="ru">Yan Z., et al. DF40: Toward Next-Generation Deepfake Detection // arXiv preprint arXiv:2406.13495. 2024.</mixed-citation><mixed-citation xml:lang="en">Yan Z., et al. DF40: Toward Next-Generation Deepfake Detection. arXiv preprint arXiv:2406.13495, 2024.</mixed-citation></citation-alternatives></ref></ref-list><fn-group><fn fn-type="conflict"><p>The authors declare that there are no conflicts of interest present.</p></fn></fn-group></back></article>
