Beyond Disclosure: Reframing Privacy as Inference Impedance in Large Language Models

Beyond Disclosure: Reframing Privacy as Inference Impedance in Large Language Models | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Beyond Disclosure: Reframing Privacy as Inference Impedance in Large Language Models Yair Oppenheim This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9006362/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Contemporary debates in AI ethics continue to frame privacy primarily in terms of disclosure, identifiability, and data access. This article argues that such a framing is no longer sufficient for embedding-based artificial intelligence systems. We introduce the Deep Personal Privacy (DPP) framework, which reconceptualizes privacy as inference impedance within high-dimensional semantic representation spaces. Rather than asking whether information has been revealed, DPP evaluates how easily sensitive attributes can be inferred from latent embeddings. We model embedding spaces as semantic transmission layers that enable indirect attribute inference through geometric alignment. Privacy risk is therefore defined in terms of cosine similarity, inference probability, and logarithmic impedance within structured inference graphs. The framework integrates ontology-driven sensitive expression mapping, representation-level perturbation mechanisms, and a multi-objective optimization procedure balancing utility and privacy. Empirical demonstrations show that DPP-based interventions reduce semantic alignment with sensitive concept prototypes and increase inference resistance while maintaining acceptable task performance. Conceptually, the framework advances a paradigm shift in AI ethics: privacy must be evaluated not only by what systems disclose, but by what they are capable of inferring. DPP thus complements existing structural and statistical privacy approaches by introducing a representation-level metric for inferential power asymmetry. Technical Communication Philosophy Information Theory Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9006362","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":599054293,"identity":"47ec5dae-fdf5-4c3a-8530-19bdcfa2bcc9","order_by":0,"name":"Yair Oppenheim","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1klEQVRIiWNgGAWjYFACxgYIzQ6iDSxI0cJzAKRFghTbJBLAJGGFBscPt334uOOOnLnk86sbfhRIMPC3dyfg13ImsXnmzDPPjC1n55Td7AE6TOLM2Q34tRxIbGbmbTucuOF2TtoNHqAWA4lcAlrOPwRrqd9w80zazT9EabkBsSXB4Ab7sdtE2SJ542EzI9AvhhvO5LDdljGQ4CHoF77z6Y8ZgCEmb3D8+LObb/7YyPG39+LXonCAARSbIJLHACTAg1c5CMg3wLWwPyCoehSMglEwCkYmAACZZlDc2hypBgAAAABJRU5ErkJggg==","orcid":"","institution":"Tel Aviv Uneversity","correspondingAuthor":true,"prefix":"","firstName":"Yair","middleName":"","lastName":"Oppenheim","suffix":""}],"badges":[],"createdAt":"2026-03-02 06:34:11","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9006362/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9006362/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104400625,"identity":"b9ace034-42e4-40a2-a1d2-0ff2f3ef02b6","added_by":"auto","created_at":"2026-03-11 12:10:32","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":587255,"visible":true,"origin":"","legend":"","description":"","filename":"BeyondDisclosureReframingPrivacyasInferenceImpedanceinLargeLanguageModels.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9006362/v1_covered_aff9e3a9-74a8-48d0-9b10-9a27549fdb5b.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eBeyond Disclosure: Reframing Privacy as Inference Impedance in Large Language Models\u003c/strong\u003e\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9006362/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9006362/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eContemporary debates in AI ethics continue to frame privacy primarily in terms of disclosure, identifiability, and data access. This article argues that such a framing is no longer sufficient for embedding-based artificial intelligence systems. We introduce the Deep Personal Privacy (DPP) framework, which reconceptualizes privacy as inference impedance within high-dimensional semantic representation spaces. Rather than asking whether information has been revealed, DPP evaluates how easily sensitive attributes can be inferred from latent embeddings.\u003c/p\u003e \u003cp\u003eWe model embedding spaces as semantic transmission layers that enable indirect attribute inference through geometric alignment. Privacy risk is therefore defined in terms of cosine similarity, inference probability, and logarithmic impedance within structured inference graphs. The framework integrates ontology-driven sensitive expression mapping, representation-level perturbation mechanisms, and a multi-objective optimization procedure balancing utility and privacy.\u003c/p\u003e \u003cp\u003eEmpirical demonstrations show that DPP-based interventions reduce semantic alignment with sensitive concept prototypes and increase inference resistance while maintaining acceptable task performance. Conceptually, the framework advances a paradigm shift in AI ethics: privacy must be evaluated not only by what systems disclose, but by what they are capable of inferring. DPP thus complements existing structural and statistical privacy approaches by introducing a representation-level metric for inferential power asymmetry.\u003c/p\u003e","manuscriptTitle":"Beyond Disclosure: Reframing Privacy as Inference Impedance in Large Language Models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-03 07:17:43","doi":"10.21203/rs.3.rs-9006362/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8df43197-a9ea-4d8a-a745-d29b893f480d","owner":[],"postedDate":"March 3rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":63743886,"name":"Technical Communication"},{"id":63743887,"name":"Philosophy"},{"id":63743888,"name":"Information Theory"}],"tags":[],"updatedAt":"2026-03-03T07:17:43+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-03 07:17:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9006362","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9006362","identity":"rs-9006362","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}