US 10110385

I'll verify this patent against live sources before summarizing.

Let me pull the actual claim language and check for any litigation or PTAB activity specific to this patent.

Let me retrieve the full text of the independent claims (1 and 4) directly.

Verification summary — US 10,110,385

I searched for the specific number 10110385 and cross-checked the record against the authoritative full text provided in the prompt. Below is what I can state with confidence versus what I cannot.

1. Bibliographic data (high confidence — consistent across Google Patents, Justia, USPTO.report, RPX)

Abstract (as reported by Google Patents/USPTO.report and mirroring the specification): A system and method for generating a signature for a document using credentials indicating an unsanctioned signing event. The system and method includes receiving a request to generate a signature of a signatory for a document, wherein the request includes a received set of credential data for a signatory, obtaining a token identifier for at least one computing device, and determining if the received set of credential data matches credentials indicating the unsanctioned signing event. The system and method further includes receiving the signature of the signatory, the document identifier, and the token identifier, and determining based at least in part on the signature, document identifier, and the token identifier, whether the received signature is associated with the unsanctioned signing event. (Note: the full text supplied in the prompt was truncated before the abstract; the above is a secondary-source reproduction, though it matches the claim language.)

2. Plain-language overview of the independent claims

Claim 1 — Computer-implemented method (verification-side)
A computer receives a signature from a signatory who has (at least) two credential sets, each mapped to a different "duress level." It also receives a document identifier derived from the document's contents, and an identity-verification identifier for a registered token that holds two private keys (each with a corresponding public key). The system then tries to re-create signatures: it generates a first candidate signature by encrypting the document identifier with the first private key alongside the first credential set; if that does not match the received signature, it generates a second candidate using the second private key and second credential set. If the second one matches, the system concludes the signatory signed at the second duress level and performs an action — one or more of: hiding information for a first account, displaying a second account, notifying security personnel, sending an alert message, recording the duress level in a data store, or repudiating transactions tied to the document.

Claim 4 — System (signing-side, token-centric)
A computing device provides services that receive a request from a signatory to sign a document, where the signatory has multiple credential sets (a first tied to a first duress level, a second to a second duress level). The service obtains a document identifier derived from document contents and a token identifier that is registered with and authorized by an identity registrar, where the token identifier includes a first and a second private key (with corresponding public keys). The service generates signature candidates from the document identifier + each credential set + the token's keys, determines which one matches, and performs an action according to the duress level of the matching (second) credential set — again from the same menu of actions (hide/display account data, notify security, message, record, or repudiate).

Claim 14 — Non-transitory computer-readable storage medium (verification/attribution)
Instructions cause a system to receive a signatory's signature, where the signatory has multiple credential sets, each mapped to a corresponding duress level; receive a document identifier derived from document contents; obtain a token identifier for a token authorized by a service provider to generate signatures and comprising a set of private keys of a public-private key scheme. It generates a first signature by encrypting the document identifier with a first private key; determines (using the signature, document identifier, token identifier, and a corresponding public key) that the first signature does not match; generates a second signature by encrypting the document identifier with the second private key and the second credential set; determines the second signature does match and is associated with a duress level indicating either a duress event or a signing event that exceeded the signatory's signing authority; and performs an action as a result.

Dependent-claim map (for context): claims 2–3 depend from 1 (claim 2 = unwilling signing; claim 3 = differing duress confidence scores); claims 5–13 depend from 4 (biometric sets incl. fingerprint/finger-movement; physiological sensors such as heart rate, pupil dilation, temperature, sweat, blood pressure; geolocation criteria; imprinting the signature as an optically scannable code; duress confidence score; acceptable credential types selectable by the signatory); claims 15–20 depend from 14 (signing-authority criteria, action selection, etc.).

Observed scope caveat: the issued claims are narrower and more machinery-specific than the specification's broad disclosure (duress passwords, fake ATM/e-mail accounts, automatic biometric duress detection). The independent claims as issued center on multiple private keys / multiple credential sets mapped to duress levels and on regenerating candidate signatures to identify which duress level was used — not on duress detection per se.

3. Litigation / CAFC 2026 docket check

No litigation, PTAB proceeding, or Federal Circuit appeal involving US 10,110,385 was found in the sources I could reach. I cannot claim this is exhaustive — I did not run a full PACER / CAFC-Rule-36 / PTAB-API docket sweep, so treat "no litigation found" as a search result, not a negative assurance.

Important de-confliction flag (per your no-auto-correct rule): my "CAFC 2026" query surfaced GoTV Streaming, LLC v. Netflix, Inc., Nos. 2024-1669, 2024-1744 (Fed. Cir. Feb. 9, 2026), which concerns US 8,478,245; US 8,989,715; and US 8,103,865 — display-tailoring patents held ineligible under §101. US 8,103,865 is a different patent from US 10,110,385, and it is unrelated to Amazon. The visual similarity (8,103,865 vs 10,110,385) is a conflation trap; I am reporting it as a distinct number and expressly not treating it as the subject patent. Likewise, Rideshare Displays, Inc. v. Lyft, Inc., No. 25-1132 (petition filed Mar. 23, 2026), which also appeared in the 2026 CAFC/§101 results, involves US 9,892,637; 10,169,987; 10,395,525; 10,559,199; and 10,748,417 — none of which is US 10,110,385.

4. Explicit uncertainty notes

• Abstract text: sourced from secondary mirrors, not the (truncated) primary text in the prompt. It is consistent with the claims, but I am flagging the sourcing.
• Claim 4's middle limitations: the exact transition language between "obtain a token identifier…" and "perform an action…" was not fully retrieved in my searches. The recitation I summarize is inferred from the parallel structure of claims 1 and 14 and the visible claim 4 language; verify against the granted claim set at https://patents.google.com/patent/US10110385 before relying on it.
• Term/expiration: "Active, expires 2035-02-17" is Google Patents' own disclaimer-backed estimate (an assumption about PTA), not a legal conclusion.
• Related-family inference (low confidence): Justia's page for US 9,866,393 ("Device for creating reliable trusted signatures") reproduces text appearing to match this patent's FIG. 9/10 discussion, suggesting a shared Amazon specification family. I did not verify a formal priority/common-ownership link and would not assert it without checking the face of that patent.
• CAFC 2026 docket coverage: my search results did not return a comprehensive CAFC docket listing; absence of a hit for US 10,110,385 should not be read as confirmation that no appeal exists.

I'll search for litigation records involving this specific patent number.

Initial searches returned only prosecution-history noise. Let me query litigation-specific sources.

Let me try patent-specific litigation databases directly.

Litigation Search Memorandum — U.S. Patent No. 10,110,385

Patent identified: U.S. Patent No. 10,110,385 B1 ("Duress signatures"), Application No. 14/580,118, filed 2014-12-22, granted 2018-10-23, assignee Amazon Technologies, Inc. (inventors Rush, Canavor, Hitchcock, Johansson, McClintock). Status per Google Patents: Active, adjusted expiration 2035-02-17.

Search date: April 26, 2026.

Result: No litigation found

I found no record of any litigation — district court, ITC, or Federal Circuit — involving U.S. Patent No. 10,110,385. I also found no IPR, PGR, CBM, or reexamination proceeding against it. Searches of litigation-aggregator content (Unified Patents-style case lists, Docket Alarm/PTAB docket pages, CourtListener/Federal Circuit opinion text, and general web/case-law indexes) returned no docket in which 10,110,385 was asserted, challenged, or adjudicated.

Because I cannot report a case, I am reporting the negative finding, with the caveats below.

Important caution — "385" false positives

Several search hits reference a "385 patent," but none of them is U.S. 10,110,385. Do not conflate these:

Each of these involves a different patent whose number merely ends in "385," or an entirely different dispute. Per your instruction to interpret identifiers literally, I am reporting only what the record shows for the exact number 10,110,385, and I am flagging the look-alikes so they are not misattributed.

Caveats on the negative finding

1. Not proof of absence. A null result from the accessible indexes is not the same as certification that no case exists. Suits asserting this patent could exist in a docket not surfaced by these searches (e.g., a recently filed complaint not yet indexed, a case where the patent is one of many asserted but not named in the caption, or a sealed/qui tam-adjacent filing).
2. Direct PACER access. I could not run a live PACER/RECAP docket-by-patent query. The most reliable confirming steps would be: (a) a PACER "Patent Cases by Patent Number" report or a Docket Navigator / Lex Machina / Unified Patents Portal patent-page lookup; (b) a Google Patents "Litigation" tab check; (c) a PTAB API query under the patent's application 14/580,118; and (d) a Federal Circuit docket search.
3. Post-grant proceedings. Litigation-adjacent activity sometimes appears only as a post-grant challenge. I found none, but a PTAB/EPO register check for the application number would confirm.

Practical assessment

10,110,385 is a 2018-granted Amazon Technologies patent (duress-credential digital signatures) with an active term to 2035. It bears the hallmarks of a defensive/portfolio asset rather than an asserted patent: no docketed assertion, no known post-grant challenge, and no news of enforcement. My best-supported conclusion is that the patent has not been litigated to date as far as publicly searchable records show, but I recommend one confirming pull from Docket Navigator or the Unified Patents patent page before relying on this in any filing, opinion, or due-diligence memo.

If you want, I can refine the search by (i) the patent's application number 14/580,118, (ii) related Amazon "identity verification token" family members, or (iii) Amazon as plaintiff/defendant in signature-authentication suits, to check for family-member or portfolio-level litigation.

I'll verify the structured data against public sources before drafting.

Let me run a few more targeted checks to confirm the absence and look for any litigation or appeal activity.

Let me run two final confirmation checks and then report.

Proceedings overview

Total AIA trial proceedings on file for US 10,110,385: 0. The USPTO Open Data Portal returns no IPR, PGR, or CBM proceeding against this patent, and my independent web checks surfaced no petition, institution decision, Final Written Decision, or Federal Circuit appeal naming US 10,110,385 (or its application 14/580,118) as the challenged patent. Because there are no proceedings, there is no breakdown by active / invalidated / sustained / settled / institution-denied status — and, critically, no claim of this patent has been canceled. The defensive posture is therefore not "hardened" and not "dead claims": US 10,110,385 stands fully intact and untested at the PTAB, with all 20 claims live and no PTAB estoppel available to help you.

Proceedings

None. There is no proceeding to enumerate. Per the structured "PTAB proceedings on file" block (USPTO ODP, most recent ingest), the PTAB trial record for US 10,110,385 is empty, and web search did not surface any un-indexed or recently-filed proceeding, so the default stated in the task instruction — "no PTAB activity on file" — controls here.

One important disambiguation: several search hits referenced a different patent with a superficially similar number, US 11,087,385 (the VB Assets / voice-assistant family in litigation with Amazon, IPR2025-00885 line). That is not US 10,110,385. Similarly, hits for US 10,536,751 and 10,028,026 are unrelated. No result tied any AIA trial to the Duress Signatures patent.

Strategic summary

Claim status. All claims of US 10,110,385 (claims 1–20, including independent claim 1 and independent claims 4 and 14 as recited in the granted text) are UNTESTED — none canceled, none confirmed, none narrowed by amendment. There is no FWD to point to, so there is no claim-level disposition to quote. For a defendant, this cuts both ways: you cannot kill an infringement theory by citing a cancellation, but you also face no adverse PTAB claim construction or estoppel-laden record that would constrain your own invalidity case.

Estoppel landscape. 35 U.S.C. § 315(e)(2) estoppel attaches only to a petitioner who obtains an FWD. With no petitioner and no FWD, no estoppel exists against anyone with respect to this patent — including Amazon as the assignee/patent owner (estoppel runs against challenging petitioners, not owners, in any event). Every prior-art ground under § 102/§ 103, and every § 112 ground available in an AIA petition, remains fully available to any defendant filing the first IPR or PGR. There is also no § 315(b) one-year bar running against anyone, no § 325(d) "same or substantially the same art" discretion risk built up from prior Office consideration, and no Fintiv-style parallel-proceeding history to work around. You would be the first mover.

Pattern signals. No repeat-petitioner pattern exists because there are zero petitioners. No defensive aggregator (Unified Patents, RPX, etc.) appears anywhere in the chain — a notable data point, since Unified and similar entities typically target broad software/authentication claims like these. No PTAB appeal activity, and (as of the sources I could reach) no Federal Circuit appeal involving this patent number. Note that US 10,110,385 has drawn forward citations (47 "Cited By" entries on Google Patents) but no backward validity challenge.

Context worth flagging to a client. This is an Amazon Technologies, Inc. patent (inventors Rush, Canavor, Hitchcock, Johansson, McClintock), filed and granted 2014-12-22 / 2018-10-23, active with an adjusted expiration of 2035-02-17. It is a defensive big-tech portfolio asset, not a troll patent, which materially reduces the probability of ever seeing an IPR. The claims are also arguably § 101-exposed (credential/biometric matching and signature generation on generic hardware) — a ground I could not verify anyone has ever raised, and which is unavailable in an IPR but available in a PGR only within 9 months of grant (long expired here) and in district court under § 282(b). That § 101 angle is likely the more productive attack than an IPR, precisely because no IPR history constrains it.

Recommended next steps

• No PTAB activity exists — say so plainly, and use the absence as intelligence. If you have received a demand or been sued on US 10,110,385, you are entering a virgin validity landscape. Verify the null result yourself before relying on it in a client memo: pull the patent's file in PTAB E2E / P-TACTS (https://ptab.uspto.gov) and confirm no AIA review is docketed, and check the patent's USPTO PatentCenter "Patent Trial" tab and the assignment record (Amazon Technologies, Inc.) for any transfer that might signal a monetization campaign.
• If you file first, you own the record. With no prior petition, you face no § 325(d) risk from prior Office consideration of your art and no estoppel from an earlier petitioner. Consider running the strongest § 103 combination and, separately, evaluating art that was before the examiner, since § 325(d) discretion is at its weakest for a first-filed petition.
• Consider whether IPR is even the right vehicle. Because the claims appear abstract-idea-implicated (a credential/biometric duress-detection scheme implemented with generic computing components), § 101 in district court may be a cleaner invalidity path than an IPR, which cannot reach eligibility. PGR is time-barred (granted 2018-10-23).
• Trial-stage milestones: N/A. There is no institution decision, no 1-year § 316(a)(11) clock running, no oral hearing, and no FWD due date, because no trial has been instituted.
• Do not cite a proceeding number. There is no FWD to link and no E2E proceeding number to cite for this patent. Any proceeding number appearing in an adversary's brief purportedly concerning US 10,110,385 should be treated as a probable citation error or a confusion with a similarly numbered patent (e.g., US 11,087,385), and verified against PTAB E2E before you respond to it.

Confidence note: The zero-proceeding finding rests on (a) the structured ODP block stating no AIA trial proceedings, and (b) targeted web searches that returned no PTAB or CAFC matter naming US 10,110,385. I could not complete a final couple of verification searches (CourtListener docket sweep and a Unified Patents-specific check) due to a tool step limit. Before filing anything that depends on this conclusion, confirm directly in PTAB E2E and via a CourtListener / PACER docket search for "10,110,385" and "14/580,118."

I'll verify the assignment record before reconstructing it. Let me run several targeted queries.

Let me pull more specific assignment and ownership data.

Assignment & Ownership Analysis — U.S. Patent No. 10,110,385 ("Duress signatures")

Application: 14/580,118 · Filed: 2014-12-22 · Granted: 2018-10-23 · Term (adjusted): 2035-02-17
Sources consulted: Google Patents legal-events timeline (authoritative copy supplied), USPTO Assignment Center / Patent Assignment Search (assignment.uspto.gov), uspto.report, Justia Patents, and general web indexes. Live reel/frame and correspondent data could not be pulled — see the data-availability note below.

⚠️ Data-availability note (read before relying on this). The Google Patents legal-events record supplied for this patent shows exactly one post-filing ownership event and does not expose a reel/frame number or a correspondent of record. My attempts to pull the Assignment Center record directly (reel/frame + correspondent) did not return those fields through the tools available in this session. I have therefore not invented a reel/frame number or an attorney name. Everything below is reported at the fidelity the accessible record supports, and the reel/frame and correspondent fields are explicitly marked not captured. This should be confirmed by a live Assignment Center lookup at https://assignmentcenter.uspto.gov/ (patent-number search) or https://assignment.uspto.gov/patent/index.html before use in any filing.

Inventors

Basis: The single recordation of 2015-06-03 conveys the inventors' interest to Amazon Technologies, Inc. and names all five of the above as assignors (Google Patents legal events: "Assignors: RUSH, DYLAN HARRIS, CANAVOR, DARREN ERNEST, JOHANSSON, JESPER MIKAEL, HITCHCOCK, DANIEL WADE, MCCLINTOCK, JON ARRON"). Because each inventor assigned to Amazon as employer, the record is consistent with all five being Amazon employees at filing. Note this is an inference from the assignment instrument, not an independent employment verification — I did not retrieve HR/SEC-level corroboration for each individual.

Unusual-pattern check: none observed. There is no evidence of any inventor departing within 12 months, no individual re-assignment of a single inventor's interest, and no separate inventor-side transfer. All five rights moved together in one instrument. This is the ordinary "big-tech employee invention" pattern, not the fragmentation that typically precedes a fire-sale. (Jesper Mikael Johansson is a prolific Amazon Technologies inventor of record on other filings, which is consistent with — but not proof of — long tenure at the time of filing.)

Original assignee

Amazon Technologies, Inc. — named on the face of the granted patent and the sole assignee in the chain of title.

• Relationship/status: Operating subsidiary within the Amazon.com, Inc. corporate family. Operating — no dissolution, bankruptcy, or acquired-and-absorbed status on the record.
• Primary line of business: E-commerce/marketplace, cloud computing (AWS), and consumer devices and services.
• Product embodying the claims: Amazon operates authentication and digital-signing infrastructure (consumer sign-in, device authentication, TPM-secured credential handling) in the same technical space the specification describes (an "identity verification token" with TPM/SGX-backed credential handling). Whether a currently shipped Amazon product or service practices these specific claims is not determinable from the record available and I will not assert it. There is no product-marketing or enforcement evidence linking this patent to a named Amazon product.
• Emphasis: Amazon Technologies, Inc. is a real operating-company assignee, not a holding/IP shell. The patent sits in a very large corporate portfolio and, per the companion litigation memo, has no docketed assertion.

Assignment timeline

Chronological list of recorded conveyances. One conveyance appears in the accessible record.

• Executed: not exposed in the record / recorded 2015-06-03 — Reel / (not captured — see data-availability note)
  • Conveyance: Assignment (original — "ASSIGNMENT OF ASSIGNORS INTEREST")
  • Assignor: Rush, Dylan Harris; Canavor, Darren Ernest; Hitchcock, Daniel Wade; Johansson, Jesper Mikael; McClintock, Jon Arron (all five, jointly)
  • Assignee: Amazon Technologies, Inc.
  • Correspondent: not captured in the accessible record. No correspondent name, firm, or address was retrievable this session. Because this is the only link in the chain, there is no recurrence to test against — the repeat-correspondent signal cannot be evaluated either way. (No correspondent is inferred or named, to avoid fabrication.)
  • Context: Ordinary employee-to-employer assignment, executed in connection with filing (recorded ~5.5 months after the 2014-12-22 filing date).

No other recorded events exist in the accessible record. Specifically, Google Patents legal events for 10,110,385 show no post-issuance assignment, no security agreement, no merger, no change of name, no license, and no release. The patent remains with Amazon Technologies, Inc.; term is active through the adjusted 2035-02-17 expiration.

Findings statement: There are records for this patent, but they reduce to a single original assignment. There is no post-issuance chain of title to reconstruct. Per the constraint that this is itself a finding: the sole recorded owner is, and has always been, the original assignee.

Timeline diagram

timeline
    title Ownership of US 10110385
    2014 : Application filed by five inventors
    2015 : Assignment recorded to Amazon Technologies Inc
    2018 : Patent granted
    2035 : Adjusted expiration

NPE / troll-pattern signals

Note on non-inference: No signal here rests on naming alone. The absence conclusions are grounded in the dated record (single 2015-06-03 recordation to Amazon; no further legal events) rather than in the fact that Amazon is a large company.

Verdict

Insufficient data — under the rubric, which defines this as "no records, or only the original assignment." That is precisely this chain: a single employee-to-employer assignment recorded 2015-06-03 (Reel not captured) conveying all five inventors' interests to Amazon Technologies, Inc., with no subsequent recorded transfer, security interest, or name change, and no assertion (see companion litigation memo). This is a benign "clean chain" result rather than a defective one: the record affirmatively shows the patent never left an operating company. If the rubric permitted a descriptive label, "operating-company-held, non-asserting" would be the practical characterization; I did not select "Defensive / non-asserting" because the chain does not terminate at a defensive aggregator as that option requires.

Confidence: The single-assignment finding and Amazon ownership are high-confidence (dated recordation to Amazon, no further legal events). The reel/frame number and correspondent of record are unresolved and should be confirmed via a live Assignment Center pull before this memo is used in any filing, opinion, or due-diligence context.

Verify at: https://assignmentcenter.uspto.gov/ (search patent number 10,110,385 or application 14/580,118) · mirror: https://assignment.uspto.gov/patent/index.html · legal events: https://patents.google.com/patent/[US10110385B1](/patent/US10110385B1)/en

Recommended confirming steps: (1) pull the exact reel/frame and correspondent for the 2015-06-03 recordation from Assignment Center; (2) confirm no later recordation exists by querying both the patent number and application 14/580,118; (3) if a broader Amazon "identity verification token" family is of interest, search the family members' assignment records — a family-level fire-sale would show up as a cluster of Amazon→third-party recordings, which the single-asset record here does not show.

I'll research the prior-art citations for U.S. Patent No. 10,110,385 specifically.

Let me retrieve the actual "References Cited" (backward citations) listed on the face of this patent.

I found the backward-citation list. Let me now retrieve the full citation details (titles, dates) for each reference and confirm against the patent's own citation table.

Prior Art Analysis — U.S. Patent No. 10,110,385 B1 ("Duress signatures")

Analyst note on scope and identifiers. Per your instruction, I limited this analysis to the exact number 10,110,385 and did not substitute similar numbers. This matters here: the earlier sections of this dossier already flagged look-alikes that other searches conflate with this patent — 10,512,385 (Samsung v. Whirlpool, IPR2026-00337), 10,702,385 (Cardiovalve v. Edwards), and 11,087,385 (VB Assets v. Amazon). None of those is this patent, and none of their citations appear below.

Patent confirmed: U.S. 10,110,385 B1, "Duress signatures," App. No. 14/580,118, filed 2014‑12‑22, granted 2018‑10‑23, Amazon Technologies, Inc., inventors Rush, Canavor, Hitchcock, Johansson, McClintock.

Citation sources used: the front‑page "Referenced Cited" list transcribed at Justia (https://patents.justia.com/patent/10110385, "Referenced Cited" section), cross-checked against Google Patents (https://patents.google.com/patent/[US10110385B1](/patent/US10110385B1)/en). Note the distinction I am preserving throughout: the "Referenced Cited" (backward) list is what you asked for, and it is not the "Cited By" (47 forward citations) list on Google Patents — that forward list includes later documents such as US 10,893,052 B1 (Facebook, "Duress password for limited account access," 2021‑01‑12) and US 10,511,447 B1 (Guardtime), which post‑date this patent and cannot be § 102 art against it.

⚠️ Contradiction check against the previously generated sections: none found. The PTAB section's statement of "all 20 claims live" and "independent claims 1, 4, and 14" is consistent with what I see here (claims 1–3 method, 4–13 system, 14–20 computer‑readable medium). One clarification worth recording: the "clauses 1–33" that appear in the specification (patents.justia.com/patent/10110385, "Embodiments of the disclosure can be described in view of the following clauses") are not the granted claims. The granted claim 1 matches clause 23 of that list, not clause 1 — do not cite clause numbers as claim numbers.

1. The decisive framing before the table

These citations did not anticipate anything — the patent issued. Every reference below was before the examiner and the application still granted on 2018‑10‑23. So a § 102 "anticipation" mapping here is necessarily hypothetical re‑application, not a record of an existing rejection. If any of these references had actually anticipated the then‑pending claims as written, there would be no U.S. 10,110,385. Any § 102 argument you build must therefore be built fresh, on the granted claim text.

The granted claim 1 (the mapping anchor), as reported by RPX insight (https://insight.rpxcorp.com/patent/US10110385B1) requires, in combination:

1. receiving a signature of a signatory associated with a first set of credential data (first duress level) and a second set of credential data (second duress level);
2. receiving a document identifier derived from document contents;
3. obtaining an identity verification identifier for a token registered with an identity registrar, where that identifier comprises a first private key and a second private key;
4. generating a first signature by encrypting the document identifier using the first private key, and finding no match;
5. generating a second signature by encrypting the document identifier using the second private key, and finding a match; and
6. performing an action in accordance with the second duress level — hiding information of a first account, displaying a second account, notifying security personnel, sending a message, storing the duress indication, or repudiating transactions.

A § 102 reference must disclose all of this (for claim 1). That is a high bar, and it explains why the art below survived prosecution.

2. U.S. patent documents cited on the face of U.S. 10,110,385

All bibliographic data (number, issue/publication date, inventor as listed) is as transcribed by Justia from the patent front page. I flag description confidence explicitly — see the caveat in § 5; I will not invent titles.

3. Non‑patent literature cited on the face of the patent

4. Assessment: what actually is the "most relevant" prior art

(a) Within the face‑of‑patent citation list, the most relevant references are the document‑identifier/signature‑generation family (#9–#11, #17–#18, #20–#21, #24, #26, #28–#29) — i.e., the electronic‑signature‑over‑a‑document‑identifier references, because they map onto claim 1 elements 2–5. The two latest pre‑filing publications, US 2014/0359722 A1 (Schultz, 2014‑12‑04) and US 2014/0372752 A1 (Sallis, 2014‑12‑18), are the ones I would pull first: they are the only cited publications issued in the final weeks before the 2014‑12‑22 filing and therefore the only ones that can be § 102(a)(1) art and be close in time to the claimed techniques.

(b) The candidate closest to the duress concept in the cited list is US 2010/0082333 A1 (Al‑Shammari, 2010‑04‑01) — multi‑credential/access art — but even on the record it does not show the two‑private‑key encryption‑of‑a‑document‑identifier‑plus‑duress‑level‑action structure of claim 1.

(c) The genuinely most on‑point duress prior art is NOT in the citation list — and that is the headline finding. The '385 front page cites no duress‑specific reference. Duress/coercion‑responsive credentials are a recognized art: the Apple v. [patentee] briefing over US 9,665,705 in IPR2022‑00602 (Federal Circuit docket 24‑1278) argues precisely about whether Mathiassen and McKeith render a "duress access / duress signal" obvious, with McKeith described as teaching a user entering a password, a unique set of clicks, or a geometric mouse pattern "concurrently with, or after a predetermined duration from, scanning his/her fingerprint" to trigger duress access (see the brief at https://archive.org/download/gov.uscourts.cafc.21085/gov.uscourts.cafc.21085.22.2.pdf). That is textbook "second credential set → duress signal" art and is the closest known teaching to claim 1 elements 1 and 6. I am citing it as context for the field, not as a front‑page citation of the '385 patent — do not label it as "cited in" US 10,110,385. Its evidentiary weight against the '385 claims requires an element‑by‑element read of McKeith against granted claim 1, which I could not complete here.

(d) Statutory availability by category:

• The 1989–2014 patents and pre‑filing publications (#1–#6, #9–#29) are the operative § 102(a)(1) (§ 102(b) for the pre‑March‑2013 items) art.
• The 2017–2018 Amazon publications and grants (#7, #8, #30, #31) post‑date the filing and are reachable only under § 102(a)(2), and only if their effective filing dates predate 2014‑12‑22; given the shared inventorship/assignee, expect the § 102(b)(2)(C) common‑ownership exception to knock them out entirely.
• The 2015/2016 printf NPL items are very likely non‑prior art by date; confirm before anyone relies on them.

5. Explicit limitations of this analysis (do not over‑read it)

1. I could not verify the titles/subject matter of most of the 31 U.S. patent documents. The source transcription lists number, date and inventor only, and my retrieval budget was exhausted before I could open each reference's front page. Where I have marked "title unverified," treat the description as my analytical inference from the citation record, not a confirmed disclosure. Do not file anything that asserts a specific reference discloses a specific claim element until each reference is read. The one title I did verify is US 5,668,928 (speech recognition), and it is plainly peripheral to the duress core.
2. The § 102 mappings in the right‑hand column are therefore a screening framework, not § 102 rejections. They state which claim elements a reference would need to disclose, and which claims those elements sit in (1–3 method, 4–13 system, 14–20 medium).
3. No reference on the face of US 10,110,385 appears to teach duress levels, duress credentials, or a duress‑triggered action, which is consistent with the earlier sections' conclusion that this patent is untested and its claims live.
4. Do not conflate "References Cited" with "Cited By." The 47 forward citations on Google Patents are not § 102 art against this patent; the Facebook duress‑password patent and the Guardtime patent belong in a freedom‑to‑operate / follow‑on analysis, not here.
5. Consistent with the earlier dossier's caveat style: two confirmation searches (a USPTO Patent Public Search "References Cited" pull for App. 14/580,118 and a per‑reference front‑page sweep) were cut off by a tool step limit. I recommend retrieving the USPTO PatentCenter "References Cited" / PTO‑892 listing for App. 14/580,118 directly, which will give the exact reference titles and the examiner's own categorization, before relying on the table in § 2.

I'll ground the § 103 analysis against the prior-art record before mapping combinations. Let me verify the key references the earlier section flagged as "title unverified," plus the duress-specific art.

The Al‑Shammari hit is a major correction to the prior-art section — it's an Arabic‑text‑stemming reference, not credential art. Let me verify a few more references and the duress‑code general knowledge before building the § 103 case.

Obviousness Analysis — U.S. Patent No. 10,110,385 B1 ("Duress signatures") under 35 U.S.C. § 103

Patent: U.S. 10,110,385 B1 · App. 14/580,118 · filed 2014‑12‑22 · granted 2018‑10‑23 · Amazon Technologies, Inc. · inventors Rush, Canavor, Hitchcock, Johansson, McClintock · adjusted expiration 2035‑02‑17.

Scope note. This section builds on, and does not repeat, the Patent summary, Prior art, PTAB, and Litigation sections above. The prior‑art universe used here is the set already assembled in the Prior art section (face‑of‑patent "References Cited," both U.S. documents and NPL), plus the duress‑specific art I verified this session. Per the standing rule, every identifier below is reported literally, and I flag both (i) an error I confirmed in the earlier prior‑art section and (ii) spelling/number conflicts rather than "fixing" them.

0. Corrections and conflict flags required before any § 103 reasoning

1. Person of ordinary skill in the art (PHOSITA)

A person having ordinary skill in the art as of 2014‑12‑22 — the '385 filing date, and the date against which every reference below must be measured — would have: (a) a bachelor's degree in computer science, electrical engineering, or a related field, plus approximately two to three years of experience in cryptographic authentication, digital signatures, and access‑control systems; or (b) less formal education offset by additional relevant experience. Such a person would be familiar with: public‑key cryptography and key enrollment (PKI), cryptographic hashing of documents, keyed hashes/HMACs, hardware security modules and the TPM specifications the '385 specification itself incorporates by reference (ISO/IEC 11889; TCG TPM Main Spec 1.2), and the long‑standing commercial practice of duress/coercion codes in physical security (silent‑alarm duress PINs, duress codes on alarm panels). This is a low‑creativity‑bar, high‑integration PHOSITA: the '385 patent's own specification concedes that the cryptographic machinery it uses (MD5, MD6, SHA‑1, SHA‑2, HMAC, salted hashes, public‑private key pairs) is conventional.

Claim‑construction anchors used in the mapping (from the granted text as reported in the earlier sections):

• "document identifier derived from document contents" → a digest/hash of the document (the specification itself says a "hash of the document 104 contents may be generated as the document identifier 106").
• "identity verification identifier … comprising a first private key and a second private key" → a registered token that holds two key pairs.
• "generating a [candidate] signature … by encrypting the document identifier using the [first/second] private key" → the textbook private‑key operation over a document digest.
• "duress level" → an ordinal classification of the credential set presented.
• "perform an action selected from the group consisting of …" → Markush‑style ("one or more of") language. This matters: proof that the art teaches any one listed alternative (e.g., "notifying security personnel," "sending a message," or "storing an indication … in a data store") supports that element.

2. The granted claims, distilled to the elements § 103 must reach

(Claim text per the earlier sections; I did not read the granted claim set first‑hand this session — the caveat in the Patent summary section applies.)

3. Reference roles

4. Ground 1 — Primary combination: A (e‑signature over a content‑derived document identifier) + B1 (McKeith/McKeeth) + B2 (Mathiassen) + C1 (TPM)

4.1 Claim 1 chart (mapping shown as what the combination must be shown to teach)

4.2 The Markush point is load‑bearing — state it explicitly

The one alternative in claim 1's action menu that no reference I have verified teaches in these terms is "hiding information of a first account / displaying a second account" (the decoy‑account idea) and "repudiating transactions." Under the claim's "one or more of" format, that gap is not fatal to a § 103 ground for claim 1: the duress reference's silent alert to law enforcement and flag signal map onto two of the enumerated alternatives. But it is fatal‑adjacent for any dependent claim that singles out decoy‑account behavior, and I did not locate pre‑2014 art for it (see § 10).

4.3 Claim 4 (system / token side)

The same combination maps, and in some respects maps better, because claim 4 is the signing‑side claim:

• "computing device that provides one or more services" with a "token identifier registered with and authorized by an identity registrar" → B2's registered portable control (memory 7/7A, sensor 5, encryption blocks 8/8B/8C, transceiver 27) + C1's registered TPM.
• "obtain a document identifier derived from document contents" → A4, A1/A2.
• "generate signature candidates … determine which matches … perform an action according to the duress level" → B2's command table + duress‑class matching, and B1's alert output.
• "first duress level / second duress level" → B2's differentiated classes (owner/administrator vs. other users; access class vs. duress class) plus B1's normal/duress input distinction.

4.4 Claim 14 (CRM)

Same mapping. Claim 14 adds the alternative "a signing event that exceeded the signatory's signing authority." That alternative finds support in B2's express owner/administrator vs. "other users" distinction and in the general privilege‑level access‑control art (D). The "set of private keys of a public‑private key scheme" element finds support in C1 (TPM key storage) and in the standard PKI practice the base references presuppose.

5. Ground 2 — Alternative primary combination: A (base e‑signature over a document digest) + B1 (McKeith/McKeeth) alone, with C1 for the key‑storage element

This is the leaner two‑reference ground, and it may be the more defensible one before an examiner or a district court because it avoids depending on B2's publication number (which I could not retrieve):

• A teaches the whole document‑signature apparatus: scan/identify the document → derive an identifier from its contents → produce and later verify a cryptographic signature using a device‑held key that is registered with a central authority.
• B1 teaches the whole duress apparatus: two different credential inputs, one of which signals duress; and an output action (silent alert to authorities; flag).
• C1 teaches that the keys live in a registered hardware module that can hold multiple keys.

Combining is a simple substitution / improvement‑in‑the‑same‑way: replace the single signing credential with a two‑credential (consent/duress) scheme, and store a second key in the same module. Nothing about the combination changes the principle of operation of either reference: the signature algorithm is unchanged; only the input set and the number of keys change.

6. Ground 3 — Dependent claims (5–13, 15–20, and 2–3)

7. Motivation to combine — why a POSITA would have done this, not merely could have

The Federal Circuit and the Board, in the duress‑access litigation, articulated the exact motivation analysis that transfers here. In the consolidated Apple v. CPC Patent Technologies proceedings (IPR2022‑00601 / ‑00602; appeal Nos. 24‑1278 et al.), the Board reasoned:

"The motivation‑to‑combine analysis is a flexible one. '[A]ny need or problem known in the field of endeavor at the time of invention and addressed by the patent can provide a reason for combining the elements in the manner claimed.' KSR, 550 U.S. at 420 … 'A person of ordinary skill is also a person of ordinary creativity, not an automaton.' Id. at 421. Thus, 'in many cases[,] a person of ordinary skill will be able to fit the teachings of multiple patents together like pieces of a puzzle.'"
— IPR2022‑00601, Final Written Decision text (archive.org/.../gov.uscourts.cafc.21085.16.0.pdf)

And, on the specific duress motivation:

"1) vehicle security was a known problem … 2) duress access was known to be desirable … 3) denying access to unauthorized users was well known … and 4) there would have been a reasonable expectation of success in the modification."
— Petitioner's motivation findings, adopted in the Board's analysis (gov.uscourts.cafc.21085.22.1.pdf)

That reasoning is field‑independent as to the duress element, and the '385 patent's own claims confirm it: the '385 duress mechanism is deliberately domain‑agnostic — claim 1 spans ATM‑style account actions and document signatures, and the specification applies the same scheme to loan applications, receipts, memorabilia, and e‑mail logins. A POSITA confronting a signature/authorization event in 2014 would therefore have looked to the duress art for the same reason the duress art existed at all.

The KSR‑/MPEP 2143 rationales that apply:

1. Combining known elements according to known methods to yield predictable results. Hashing a document → signing the hash with a device‑held private key → adding a second credential parameter to the signed input. Each step is a known technique; the result (two different, verifiable signatures for the same document depending on which credential was entered) is precisely what the design demands, with no change in the principle of operation of any element.
2. Simple substitution of one known element for another. B1 substitutes a duress input for the normal input, and a duress output for the normal output, while leaving the underlying access decision intact — the Board's own characterization of the McKeith modification ("granting access to the car door locks and issuing a security alarm … when the user is under duress").
3. Improving a similar device/technique in the same way. Duress signaling had already been used to improve lock/access systems (B2's duress class; B1's silent alert). Applying it to a signature system improves the same thing — the security of a credential‑based transaction — in the same way.
4. Design incentive rooted in the coercer's presence. B1 teaches the disarmingly simple, decisive requirement: the duress path must look like the normal path ("grant access but issue a silent alert"). Once a signal must be indistinguishable to an observer, the combination with a signature becomes a natural design move, because a signature is exactly the artifact that leaves a verifiable record. Two credentials → two distinct but externally identical‑looking signatures is the direct engineering consequence.
5. "Obvious to try" with a finite, small, predictable solution set. The design space is essentially binary (one key or two; two credentials or one; alert or no alert), and the references point squarely at one answer.
6. Statements in the specification against interest. The spec describes the signature as the output of a conventional hash ("MD5, MD6, SHA‑1, SHA‑2," HMAC), the token number as "a private key of a public‑private key pair," the DID as optionally "a hash of the document 104 contents," and the key storage as a TPM/SGX/SIM environment — all of which are the prior art. The only asserted novelty is which values are fed in and what is done with the classification, not any cryptographic advance.

A candid note on the '385's own claimed advantage. The spec touts that a signatory "can signal his/her lack of consent … without revealing this lack of consent to observing parties." That is the defining property of a duress code, and it is expressly taught by B1. It is therefore an expected, not an unexpected, result.

8. Reasonable expectation of success

Every claimed element is implementable at the software level with techniques the references and the applicant's own incorporated standards already supply: add a second input to a hash; provision a second key pair in a TPM‑class module; test credential class A, then class B; branch on the result. Nothing requires new hardware, new mathematics, or an unpredictable physical result. Under KSR's "predictable results" rubric, and the Board's express finding that the McKeith modification "only requires simple programming techniques … within a POSITA's expertise," the expectation of success is high.

9. Anticipated counter‑arguments and how I would meet them

10. Gaps I could not close (do not over‑read this analysis)

1. No pre‑2014 reference verified for decoy/"second account" behavior or transaction repudiation. The action menu's alert/store alternatives are covered by B1; the decoy‑account alternative is not. If decoy behavior matters (dependent claims, or claim 14's action selection), a dedicated search is required. Note the follow‑on US 10,893,052 B1 (Facebook, "Duress password for limited account access," 2021‑01‑12) is a forward citation and post‑dates the '385 filing — not prior art.
2. No verified pre‑2014 reference for numeric "duress confidence scores" (claim 3).
3. Titles/subject matter of roughly 25 face‑cited U.S. publications remain unverified (the prior‑art section's caveat stands; correction 0.1 shows why it matters). Pull the PTO‑892 for App. 14/580,118 from USPTO PatentCenter and read each reference's front page before filing anything.
4. Mathiassen's publication number was not retrieved, and I have not guessed it. Cite it by name and by its role in the IPR record until the number is confirmed.
5. The duress‑art line of authority is contested and still being litigated — the Board found the Mathiassen+McKeith+Anderson combination unpatentable in IPR2022‑00601/‑00602, found divergent results in the co‑pending IPR2022‑01006, and appeal/Supreme Court petition activity continues (see the consolidated appeal Nos. 24‑1278 et al. and Supreme Court docket petition 25‑375). Use its motivation‑to‑combine reasoning as persuasive support, not as binding authority.

11. Secondary considerations (objective indicia)

On the record assembled across this dossier there is no evidence of commercial success, long‑felt but unmet need, failure of others, unexpected results, copying, or licensing attributable to the '385 claims. The patent is unasserted (per the Litigation and PTAB sections) and appears to be a defensive portfolio asset. Absent a nexus, secondary considerations do not rebut the prima facie case above. The '385's own specification supplies the strongest anti‑nexus point: the duress‑code concept it describes was a known commercial practice in physical security long before 2014.

12. Bottom line

• A prima facie § 103 case exists for claims 1, 4, and 14 on the combination {base electronic‑signature reference (A1/A2, backed by A4 for content‑derived document identification)} + {McKeith/McKeeth (US 2005/0022005 A1 / US 10,311,221 B2) as the duress reference} + {a registered multi‑key/multi‑authority token, taught by Mathiassen and/or the TPM specifications the applicant incorporated}. Ground 2 (without Mathiassen) is leaner and may be procedurally safer.
• The single most vulnerable element is "two private keys in one registered token." Everything else — content‑derived document identifiers, private‑key signing of a digest, re‑generation‑and‑compare verification, duress credential sets, and duress‑triggered alerts — is squarely taught or plainly conventional.
• The Markush "one or more of" phrasing of the action element is what makes the ground work, because two of the listed actions ("notifying security personnel," "sending an alert message," "storing an indication of the duress level") are directly taught by the duress reference.
• Flag (required): the earlier prior‑art section's characterization of US 2010/0082333 A1 (Al‑Shammari) as credential art and "the closest … to the duress concept" is wrong — it is an Arabic text‑processing publication and must be removed from any combination.
• Practical positioning. Since no IPR/PGR has ever been filed against this patent (per the PTAB section), a first‑filed petition faces no § 325(d) "same art" discretion risk and no petitioner estoppel. But because IPR cannot reach § 101 and the claims read on generic‑hardware credential matching, the eligibility attack in district court under § 282(b) — and the § 103 ground above — should be evaluated together rather than in isolation.

Confidence: high on the duress references (B1 verified directly; B2 verified through the Board/brief record) and on the Al‑Shammari correction; moderate on the base‑reference identification, because the face‑cited e‑signature publications' subject matter remains unverified; low on the two‑private‑keys element and on decoy‑account art. Confirm every reference's front page and the PTO‑892 before relying on this in a filing, opinion, or diligence memo.