{"id":41280,"date":"2026-06-06T05:27:05","date_gmt":"2026-06-06T03:27:05","guid":{"rendered":"https:\/\/www.graviton.at\/letterswaplibrary\/how-deepfake-detection-models-perform-across-social-media-platforms\/"},"modified":"2026-06-06T05:27:05","modified_gmt":"2026-06-06T03:27:05","slug":"how-deepfake-detection-models-perform-across-social-media-platforms","status":"publish","type":"post","link":"https:\/\/www.graviton.at\/letterswaplibrary\/how-deepfake-detection-models-perform-across-social-media-platforms\/","title":{"rendered":"How Deepfake Detection Models Perform Across Social Media Platforms"},"content":{"rendered":"<p><!-- SC_OFF --><\/p>\n<div class=\"md\">\n<p>When images are run through social media platforms, they are resized, re-encoded, and pushed through the platform&#8217;s codec. In assessing a deepfake detector model, it&#8217;s important to ensure the model remains robust across real world platforms. <\/p>\n<p>I built a dataset of varied image formats that mimic the image adjustments made by these popular platforms and tested some open source models on it.<\/p>\n<p><a href=\"https:\/\/huggingface.co\/datasets\/danb21\/social-media-robustness-sdxl-instantid\">Dataset<\/a>, <a href=\"https:\/\/babalolad.substack.com\/p\/how-deepfake-detectors-perform-under?r=2kub4w\">Model Results<\/a><\/p>\n<\/div>\n<p><!-- SC_ON -->   submitted by   <a href=\"https:\/\/www.reddit.com\/user\/Tasty_Pressure_5618\"> \/u\/Tasty_Pressure_5618 <\/a> <br \/> <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1ty4y9v\/how_deepfake_detection_models_perform_across\/\">[link]<\/a><\/span>   <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1ty4y9v\/how_deepfake_detection_models_perform_across\/\">[comments]<\/a><\/span><\/p><div class='watch-action'><div class='watch-position align-right'><div class='action-like'><a class='lbg-style1 like-41280 jlk' href='javascript:void(0)' data-task='like' data-post_id='41280' data-nonce='78a28cb461' rel='nofollow'><img class='wti-pixel' src='https:\/\/www.graviton.at\/letterswaplibrary\/wp-content\/plugins\/wti-like-post\/images\/pixel.gif' title='Like' \/><span class='lc-41280 lc'>0<\/span><\/a><\/div><\/div> <div class='status-41280 status align-right'><\/div><\/div><div class='wti-clear'><\/div>","protected":false},"excerpt":{"rendered":"<p>When images are run through social media platforms, they are resized, re-encoded, and pushed through the platform&#8217;s&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[85],"tags":[],"class_list":["post-41280","post","type-post","status-publish","format-standard","hentry","category-datatards","wpcat-85-id"],"_links":{"self":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/41280","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/comments?post=41280"}],"version-history":[{"count":0,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/41280\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/media?parent=41280"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/categories?post=41280"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/tags?post=41280"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}