{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/detecting-adversarial-image-examples-in-deep","title":"Detecting Adversarial Image Examples in Deep Networks with Adaptive Noise Reduction","arxiv_id":"1705.08378","date":"2017-05-23","proceeding":null,"authors":["Bin Liang","Hongcheng Li","Miaoqiang Su","Xirong Li","Wenchang Shi","Xiao-Feng Wang"],"abstract":"Recently, many studies have demonstrated deep neural network (DNN)\nclassifiers can be fooled by the adversarial example, which is crafted via\nintroducing some perturbations into an original sample. Accordingly, some\npowerful defense techniques were proposed. However, existing defense techniques\noften require modifying the target model or depend on the prior knowledge of\nattacks. In this paper, we propose a straightforward method for detecting\nadversarial image examples, which can be directly deployed into unmodified\noff-the-shelf DNN models. We consider the perturbation to images as a kind of\nnoise and introduce two classic image processing techniques, scalar\nquantization and smoothing spatial filter, to reduce its effect. The image\nentropy is employed as a metric to implement an adaptive noise reduction for\ndifferent kinds of images. Consequently, the adversarial example can be\neffectively detected by comparing the classification results of a given sample\nand its denoised version, without referring to any prior knowledge of attacks.\nMore than 20,000 adversarial examples against some state-of-the-art DNN models\nare used to evaluate the proposed method, which are crafted with different\nattack techniques. The experiments show that our detection method can achieve a\nhigh overall F1 score of 96.39% and certainly raises the bar for defense-aware\nattacks.","url_abs":"http://arxiv.org/abs/1705.08378v5","url_pdf":"http://arxiv.org/pdf/1705.08378v5.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"detecting-adversarial-image-examples-in-deep","repo_url":"https://github.com/OwenSec/DeepDetector","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"detecting-adversarial-image-examples-in-deep","repo_url":"https://github.com/son-of-win/DeepDetectorDefense","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"quantization","task_name":"Quantization"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1705.08378","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}