[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85753-en":3,"doc-seo-85753-105":29,"detail-sidebar-cat-0-en-105":83},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85753,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",8,"Research & Report","What You Train Is What You Get Gender Bias Training Composition and Post Hoc Mitigation in Audio Deepfake Detection","Audio deepfake detection systems decide whether speech is authentic or artificially generated, yet high aggregate accuracy can hide large performance gaps across demographic groups. This study analyzes gender bias using the ASVspoof5 dataset with a controlled custom split to isolate training composition effects. Attack-specific ResNet18 models are trained on nine gender-composition sets and evaluated with LogSpectrogram and WavLM-Base+ features plus six post-hoc threshold calibration methods. Training composition predicts bias direction; underrepresented genders perform worse at test time. WavLM-Base+ yields 3.0–4.3× larger gaps than LogSpectrogram, while calibration cannot change the Equal Error Rate gap (1.317 pp).","What You Train Is What You Get: Gender Bias, Training Composition, and Post-Hoc Mitigation in Audio Deepfake Detection  \nAishwarya R. Fursule, Vamshi Nallaguntla, Shruti Kshirsagar, and Anderson R. Avila  \narXiv :2607 .0989 1v 1 [ cs . SD] 10 Jul 2026  \nAbstract—Audio deepfake detection models determine whether speech is genuine or artificially generated, but high overall accuracy can mask substantial performance disparities across demographic groups. In this work, we investigate gender bias in audio deepfake detection using the ASVspoof5 dataset. We use ASVspoof5 under a controlled custom split designed to isolate gender-composition effects. We train attack-specific models on nine training sets with different gender compositions, ranging from female-only to male-only. We use a ResNet18 classifier with LogSpectrogram and WavLM-Base+ features, and we evaluated six post-hoc threshold calibration methods. Experimental results show that training data composition strongly predicts bias direction, with the underrepresented gender performing worse attest time. WavLM-Base+ features are shown to produce gender performance gaps 3.0 to 4.3 times larger than LogSpectrogram under identical training conditions, and balanced training is found to reduce LogSpectrogram bias but leave WavLM bias largely intact. Moreover, all six calibration strategies, including Oracle calibration with full test-set label access, leave the Equal Error Rate gap unchanged at 1.317 pp, confirming that threshold adjustment cannot correct underlying score distribution disparities. Overall, these findings suggest that gender fairness in audio deepfake detection must be addressed at training time, as posthoc methods can only partially mitigate the resulting disparities.  \nIndex Terms—Audio deepfake detection, gender bias, training data composition, threshold calibration, spoofing attacks, trustworthy AI  \nI. INTRODUCTION  \nSPEECH can no longer be reliably trusted as an indicator  \nof human identity [1] . Advances in neural text-to-speech synthesis, voice conversion, and adversarial perturbation now enable generation of audio that can convincingly mimic human speech and deceive listeners [2], [3] . The research community has responded with a range of countermeasures, from classical Gaussian Mixture Model classifiers to end-to-end architectures such as RawNet2 [8], graph attention networks such as AASIST [9], and self-supervised representations including WavLM [38], all benchmarked through the ASVspoof challenge series [7], [34]–[36] . However, strong average accuracy does not guarantee equitable performance across demographic groups [29] as a model may achieve a low aggregate Equal Error Rate (EER) while consistently failing one gender on specific attack types, implying that some users receive less protection based on gender alone.  \nA. Fursule, V. Nallaguntla, and S. Kshirsagar are with the School of Computing, Wichita State University, Wichita, KS, USA (e-mail: [axfursule@shockers.wichita.edu](axfursule@shockers.wichita.edu); [vxnallaguntla@shockers.wichita.edu](vxnallaguntla@shockers.wichita.edu); [shruti.kshirsagar@wichita.edu](shruti.kshirsagar@wichita.edu)).  \nA. R. Avila is with the Institut national de la recherche scientifique (INRS–EMT), Montreal, QC, Canada, and also with the INRS-UQO Mixed Research Unit on Cybersecurity, Gatineau, QC, Canada (e-mail: ander[son.avila@inrs.ca](son.avila@inrs.ca)) .  \nPrior work has established that this failure mode exists but has not resolved its cause, particularly with respect to how attack type, speaker gender, and feature representation interact [4]–[6] . Detection systems trained on female voices outperform those trained on male voices [12], attributed to spectral artifacts in synthesized female speech. A fairness audit of six detection systems [20] found the opposite: male speakers face higher false positive rates regardless of architecture. A third study [15] attributes disparities to score distribution shift between ","cbCaiiLO00t4HcEU","https://ap.wps.com/l/cbCaiiLO00t4HcEU","pdf",4243495,1,15,"English","en",105,"# Introduction\n## Speech authenticity and detection advances\n## Prior work and fairness contradictions\n# Method overview\n## Controlled dataset split and gender-composition training\n## Models, features, and post-hoc calibration","[{\"question\":\"Why can high accuracy in audio deepfake detection still be unfair across genders?\",\"answer\":\"Because aggregate Equal Error Rate can remain low while performance consistently fails for one gender on specific attack types, meaning some users receive less protection based on gender alone.\"}]",1784206026,38,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":78,"head_meta":80,"extra_data":82,"updated_unix":27},"what-you-train-is-what-you-get-gender-bias-training-composition-and-post-hoc-mitigation-in-audio-deepfake-detection","",{"@graph":35,"@context":77},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/what-you-train-is-what-you-get-gender-bias-training-composition-and-post-hoc-mitigation-in-audio-deepfake-detection/85753/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71],{"name":72,"@type":73,"acceptedAnswer":74},"Why can high accuracy in audio deepfake detection still be unfair across genders?","Question",{"text":75,"@type":76},"Because aggregate Equal Error Rate can remain low while performance consistently fails for one gender on specific attack types, meaning some users receive less protection based on gender alone.","Answer","https://schema.org",{"og:url":51,"og:type":79,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":81,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":84},[85,89,93,97,102,107,112,115,120,123,127],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":86,"show_sort_weight":87,"slug":88},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":90,"show_sort_weight":91,"slug":92},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Exam",70,"exam",{"id":98,"doc_module":4,"doc_module_name":45,"category_name":99,"show_sort_weight":100,"slug":101},5,"Comic",60,"comic",{"id":103,"doc_module":4,"doc_module_name":45,"category_name":104,"show_sort_weight":105,"slug":106},6,"Technology",50,"technology",{"id":108,"doc_module":4,"doc_module_name":45,"category_name":109,"show_sort_weight":110,"slug":111},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":113,"slug":114},30,"research-report",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":45,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":98,"slug":130},19,"General","general"]