US 9031883

I'll verify key bibliographic details and check for any 2026 litigation activity before summarizing.

Let me pull the actual claim set and check for any 2026 litigation activity.

I have the specification. Now let me retrieve the actual claim set, since the provided text stops before the claims.

Search scope note

I ran searches against public patent databases (Google Patents, USPTO.report, Justia, Patents-Review, FPO), general web, and district-court docket aggregators (Justia Dockets, CourtListener, PacerMonitor, UniCourt, govinfo). I did not find any CAFC 2026 docket, IPR, or district-court case that names U.S. Patent No. 9,031,883 or asserts it. The 2026 Meta-related patent suits that surfaced (Haptix Solutions, GAEA, Datascription, Mullen Industries, NovaCloud Licensing, EasyWeb Innovations, Solos Technology) do not identify '883 in the results I could see. I cannot independently confirm the absence of litigation — I only note that no case surfaced in my searches. Treat the litigation point as unverified.

One caveat on the claim overview below: the authoritative text I was given for US 9,031,883 stops at the end of the specification and does not include the claims section, and my searches did not return the verbatim claim language. So the claim descriptions below are reconstructed from the patent's own SUMMARY section (which tracks the claims) and the abstract. I flag where I am inferring.

---

Bibliographic summary — US 9,031,883 B2

Family note (same title, same inventors): US 9,558,279 B2 (continuation, app. 14/668,030, filed 2015-03-25, granted 2017-01-31); US 10,740,419 B2 (continuation, app. 15/387,571); US 11,636,166 B1 (continuation, app. 16/919,083). '883 is the root filing of this chain. (Do not confuse with the similarly-titled but distinct Facebook patent US 10,776,437 B2, Time-Window Counters for Search Results on Online Social Networks, Gupta et al. — different inventors and claims.)

Abstract (as issued)

"To allow for tracking events and classifying assets within a social networking system. A time series of occurrences of an event type associated with at least one asset is generated. A first signal value and a second signal value is determined based on the time series. The at least one asset is classified based on comparison of the first signal value and the second signal value. In an embodiment, the time series is based on at least one time window including time intervals. In an embodiment, counters to determine a number of occurrences of an event type are associated with the time intervals. In an embodiment, each of the counters are incremented upon occurrence of the event type associated with the at least one asset during an associated time interval."

Plain-language overview of the independent claims

Based on the SUMMARY section (which mirrors the claim set). Verbatim claim text was not available to me — confirm against the printed claims before relying on this.

The disclosure is directed to counting recurring user "events" (e.g., sharing, clicking, hiding a URL) about an "asset" (a URL, user, IP address, content item, message), storing those counts in rolling time buckets, deriving numeric "signal values" from the resulting time series, and feeding those signals into a classifier — chiefly to catch spam versus genuine popularity.

• Independent method claim (claim 1 in the issued patent's scheme). A computer-implemented method in which a time series of occurrences of an event type tied to an asset is generated; a first signal value and a second signal value are determined from that time series; and the asset is classified based on a comparison of the two signal values. The time series is built on a time window subdivided into time intervals, with a counter per interval incremented each time the event occurs during that interval.
  • Uniqueness variant: in a dependent/related aspect, at least one of the two signal values is "based on uniqueness" — e.g., counting distinct users who shared a URL rather than raw share counts, so 100 shares by one user can be distinguished from 100 shares by 100 users.
  • Signals via integration: a signal value may be obtained by computing an integral/area under the curve of the time series over a chosen period (e.g., f₁ = shares in the last hour; f₂ = shares in the last 3 hours; g₁, g₂ = the "unique user" analogues).
• Independent system/apparatus claim. The same functionality expressed as a system (one or more processors plus memory/computer-readable storage holding instructions), i.e., generating the event time series, determining the multiple signal values, and classifying the asset based on their comparison.
• Independent computer-readable medium claim. The same functionality as a non-transitory computer-readable medium storing instructions that, when executed, cause the above steps to be performed.
• Classification mechanics (dependent and/or alternative aspects): the classifier may be a decision tree or machine-learning model that produces a score to which the asset is assigned; classification may place the asset in a spam, popularity, or ranking category; and a system policy may be modified based on the resulting classification (e.g., escalating verification challenges for a URL suspected of spam).
• Buffer implementations (dependent aspects):
  • Circular buffer — each element is a time interval; the counter in the oldest interval is reset and reused for the newest interval (the buffer overwrites the least-recent interval).
  • Linear buffer with exponentially-decaying intervals — values roll from the just-expired counter into the next, progressively longer counter (e.g., 1 min → 2 min → 4 min → 8 min), preserving fine-grained detail for recent events while aging coarse aggregates; the oldest interval may be fixed or expanding.
  • Linear buffer with non-decreasing (fixed) intervals — a permanently retained per-interval counter, enabling comparisons of like periods across time (e.g., a Sunday noon spike vs. previous Sundays' noon), at higher storage cost.
• Real-time limitation: generating, determining, and/or classifying may be performed in real time (or asynchronously/batch).

Bottom line: the asserted core of the '883 family is not the raw idea of counting events, but the specific combination of (a) bucketed, windowed counters (with circular or decaying/fixed linear buffer implementations), (b) derivation of plural signal values from the resulting time series (optionally with a uniqueness dimension), and (c) automated classification of the asset by comparing those signals — targeted at spam detection in a social network.

I'll search for litigation involving US patent 9031883 across patent litigation sources.

Let me check litigation-specific databases and Unified Patents.

Let me check the Unified Patents litigation portal and other patent-specific sources directly.

Litigation Search Report — U.S. Patent No. 9,031,883

Patent identified (literal match only): U.S. Patent No. 9,031,883 B2 — "Systems and methods for event tracking using time-windowed counters." Application No. 13/631,523; filed 2012-09-28; granted 2015-05-12. Original assignee Facebook, Inc.; current assignee Meta Platforms, Inc. Inventors: Bhalchandra Suresh Agashe, Gil Shklarski, Christopher Alexander Stein, Nickolay Vladimirov Tchervenski. (Source: https://patents.google.com/patent/[US9031883B2](/patent/US9031883B2)/en; https://uspto.report/patent/grant/9,031,883)

Finding: No known patent litigation found involving U.S. Patent No. 9,031,883

After searching patent-litigation-oriented sources (Unified Patents litigation coverage, PTAB/PTACTS records, general docket references, and litigation-news coverage), I found no district court case, ITC action, or other infringement litigation naming U.S. Patent No. 9,031,883 as an asserted patent. Specifically:

Important caveats

1. Search limitations. I was unable to directly query PACER, CourtListener/RECAP, or the Unified Patents litigation caselist in this session; my checks relied on publicly indexed web sources. It is possible that a litigation exists that is not surfaced in those indexed sources. Free/web sources are not a complete substitute for a PACER or Docket Navigator query.

2. No similar-number substitution. Per your instruction, I did not treat results for numerically similar patents (e.g., 9,031,865; 7,535,890; 8,995,433; 9,631,886) as relevant. Those appeared in search results but are different patents and are excluded here. Facebook/Meta is frequent litigation and IPR-party in unrelated Uniloc-family matters (e.g., Uniloc 2017 LLC v. Facebook, Inc., CAFC No. 19-2162), but none of those involved '883.

3. Assignee-level context. Meta Platforms (formerly Facebook) is a large, active patent holder, and this patent sits in a family (relatives include US 9,558,279; US 10,740,419; US 11,636,166 per the Google Patents record). I found no indication that '883 itself has been asserted.

Recommendation

For a definitive answer, query:

• PACER / CourtListener RECAP for "9,031,883" or "9031883" in case dockets;
• Unified Patents litigation portal (portal.unifiedpatents.com/litigation) and the PTAB for any IPR/PGR involving this patent (none appeared in my searches);
• Docket Navigator or Lex Machina for a comprehensive assertion history.

If you have a specific court, party, or date range you suspect, I can refine the search. Based on the evidence available to me as of this search, the answer is: no litigation is known involving U.S. Patent No. 9,031,883.

I have interpreted the identifier "9031883" literally as U.S. 9,031,883 and have not auto-corrected it. Where source data and my training conflicted, I deferred to the searched sources.

I'll verify the structured block's "no proceedings" default against live sources before drafting.

The structured block says no proceedings; let me probe whether any party ever petitioned against the '883 patent (e.g., Sound View Innovations campaigns) or whether it was asserted in district court.

Let me do a final targeted check across PTAB dockets, the patent family, and any litigation assertion of the '883 patent.

PTAB / AIA Trial Landscape — U.S. Patent No. 9,031,883

Patent: Systems and methods for event tracking using time-windowed counters (Agashe, Shklarski, Stein, Tchervenski)
App. No. 13/631,523 · Filed 2012-09-28 · Granted 2015-05-12 · Adjusted expiration 2032-10-25
Assignee: Meta Platforms, Inc. (originally Facebook, Inc., change of name 2022-01-27)

---

Proceedings overview

Total AIA trial proceedings on file: 0. Breakdown: 0 active, 0 with claims invalidated, 0 with claims sustained, 0 settled, 0 institution-denied — the structured USPTO Open Data Portal block reports no AIA trial proceedings for US 9,031,883 as of the most recent ingest, and I could not surface any IPR, PGR, or CBM petition against this patent in live search either. The defensive posture this gives a defendant is neither "hardened by survival" nor "gutted by cancellation" — the claims are simply UNTESTED at the PTAB, and there is no claim-level disposition, no FWD to cite, and no § 315(e)(2) estoppel to exploit.

No proceeding records to populate

The requested per-proceeding template (### {PROCEEDING_NUMBER} — {Petitioner} v. {Patent Owner}) is intentionally left empty. I will not manufacture a proceeding number, petitioner, panel, or claim-level outcome, because none exists in the record:

• Proceeding number: none on file — no IPR20xx-xxxxx, PGR20xx-xxxxx, or CBM20xx-xxxxx is associated with this patent.
• Petitioner / Patent Owner: N/A
• Type: N/A (note: PGR is unavailable for this patent — it has a pre-AIA effective filing date of 2012-09-28; CBM is unavailable because the transitional program sunset on 2020-09-16)
• Filed / Status / Judge panel / Grounds / Institution decision / FWD / Settlement / Appeal: N/A — no record
• Defensive value: see the strategic summary below; the operative fact is the absence of any administrative validity challenge.

Verification note: the canonical source remains the USPTO Open Data Portal / PTAB E2E dockets. Anyone relying on this should re-run the ODP query and a PTAB E2E patent-number search at the time of filing, because a recently filed petition can lag the ODP ingest. My searches on 2026-09-29 surfaced none.

---

Strategic summary

Claims CANCELED: none. Claims SUSTAINED: none. Claims adjudicated in any forum: none. Every claim of the '883 patent is UNTESTED. I did not find the patent asserted as a patent-in-suit in any district court action either — notably, it was not among the seven patents Sound View Innovations asserted against Facebook in Sound View Innovations, LLC v. Facebook, Inc., No. 1:16-cv-00116 (D. Del.) (the '845, '371, '181, '786, '486, '593, and '860 patents), nor in Sound View's later waves against Hulu, Twitter, Fidelity, Walmart, Delta, and the media defendants. That is a meaningful signal: this is a Facebook/Meta-originated portfolio patent that has never been asserted and has never been attacked.

Estoppel landscape: § 315(e)(2) estoppel is a non-issue here — no petitioner has ever reached an FWD on this patent, so no party (or privy) is barred from raising any § 102/§ 103 ground. A defendant today has the full universe of prior art available for an IPR petition, unconstrained by estoppel. The constraints that do matter are procedural, not estoppel-based: (i) the § 315(b) one-year clock, which starts the day a complaint alleging infringement of the '883 patent is served; (ii) § 315(a)(1), which bars an IPR if the defendant first files a declaratory-judgment action of invalidity; and (iii) the discretionary-denial doctrine, which has tightened considerably. Because IPR is limited by § 311(b) to anticipation and obviousness on patents and printed publications, a defendant wanting § 101 or § 112 must litigate those in district court — no AIA window is open for them (PGR unavailable for this pre-AIA patent; CBM sunset).

Pattern signals: there is no serial-petitioner pattern, no defensive-aggregator involvement (no Unified Patents or similar filing), and no patent-owner PTAB appeal history to read, precisely because there is no proceeding. The relevant portfolio signal is the continuation chain: the Google Patents record shows later U.S. applications claiming priority to the '883 application, including US 14/668,030 (US 9,558,279), US 15/387,571 (US 10,740,419), and US 16/919,083 (US 11,636,166). A defendant should confirm whether the asserting entity is asserting the '883 patent alone or the family, since invalidating the '883 claims does not touch the separate claims of the continuations, and each has its own independent validity posture.

Time-sensitive caveat: the PTAB is materially different from the 2015–2024 baseline. Director Squires' 2025-10-16 memorandum moved institution decisions to the Director, and the 2025-10-15 NPRM proposed barring IPR against patents that have already survived a validity challenge plus a stipulation requirement on parallel § 102/§ 103 arguments. Reported institution rates have collapsed. I do not know with confidence whether that NPRM was finalized as of 2026-09-29, or how a pending petition against an 11-year-old patent would fare under the agency's "settled expectations" practice — treat that as a live diligence item, not a settled rule. (Willkie client alert; BSA letter, 2026-09-02)

---

Recommended next steps

1. Do not rely on a PTAB win that does not exist. There is no FWD to cite and no canceled claim to anchor a non-infringement or invalidity position. Re-verify the zero count against USPTO PatentCenter and the PTAB E2E docket immediately before any filing, using both the patent number (9,031,883) and the application number (13/631,523).
2. Build the invalidity case from scratch, on § 102/§ 103. Because no IPR has been filed, the full prior-art record is open and there is no estoppel. Map the time-windowed counter / time-series / asset-classification limitations (circular buffer, linear buffer with exponentially-decaying intervals, signal-value comparison, decision tree or machine-learning classifier) against pre-2012 social-networking event-logging and rate-limiting art. The abstract frames the core claim as generating a time series of event occurrences, determining first and second signal values, and classifying an asset by comparing them — that is the limitation set to attack.
3. If you are served, calendar the § 315(b) deadline immediately. One year from service of the complaint; and do not file a DJ of invalidity first, or § 315(a)(1) forecloses the IPR entirely.
4. Split your § 101/§ 112 theories into the district court track now. The claims (event tracking, spam classification, rankings) present a conventional Alice / eligibility argument that no AIA vehicle can hear for this patent. Budget for it as a court-only defense.
5. Confirm the asserted family scope. Ask the patent owner for an infringement contention chart identifying exactly which patent numbers in the '883 family are asserted. Check whether US 9,558,279, US 10,740,419, or US 11,636,166 are in play, and note that their much later priority dates (2015, 2016, 2020) may make them PGR-eligible where the '883 patent is not.
6. If you are considering filing an IPR, run the discretionary-denial analysis before spending on the petition. Given the current Director-level institution practice and the patent's age, expect a hard look on discretionary grounds; a petition filed without addressing those factors up front is unlikely to be instituted.

Bottom line for a defendant: the absence of PTAB activity is itself the signal here. This patent has never been asserted in the litigation I could identify and has never drawn a petition, which is consistent with a defensive portfolio asset rather than an active monetization tool. But that cuts both ways — you get no free claim cancellations, no estoppel leverage, and no panel findings to lean on. The validity fight is a blank slate, and any IPR you file today faces an institution environment far more hostile than the one this patent was granted into.

I'll research the assignment record for US 9031883 before reconstructing the chain.

Ownership & Assignment Analysis — US Patent 9,031,883 B2

Patent: Systems and methods for event tracking using time-windowed counters
App. No.: 13/631,523 · Filed: 2012-09-28 · Granted: 2015-05-12
Assignee on the face of the grant: Facebook, Inc.

Sourcing note / limitation. The Google Patents legal-events record (authoritative text supplied) contains two recorded post-filing events, but does not expose reel/frame numbers. My live lookups against USPTO Assignment Center / uspto.report did not return the patent-specific reel/frame entries within this session, so I cannot state them verbatim. Where a reel/frame is shown below it is for the trademark counterpart of the same 2021 name change, clearly labeled as such, and must not be cited as the patent's reel/frame. Treat every reel/frame field on this patent as unverified until pulled directly from Assignment Center.

---

Inventors

All four inventors assigned their entire interest to Facebook, Inc. in the recorded 2013 instrument (Google Patents legal events: "ASSIGNMENT OF ASSIGNORS INTEREST … Assignors: SHKLARSKI, GIL, AGASHE, BHALCHANDRA SURESH, TCHERVENSKI, NICKOLAY VLADIMIROV, STEIN, CHRISTOPHER ALEXANDER").

Pattern notes:

• This is a four-inventor team filing typical of an in-house integrity/anti-spam engineering group at Facebook — the patent is squarely about spam detection and asset scoring (spec ¶ describing classifying a URL as spam based on unique-user sharing counts). Nothing anomalous: it is a run-of-the-mill employee-invention assignment, not a solo-inventor shell-filing.
• No evidence of an inventor-departure-before-fire-sale pattern. I found no assignment records, court filings, or SEC disclosures indicating that the inventors' departure was coupled to any transfer of this patent. The "all inventors left within 12 months" tell is not present / not verifiable from the assignment record. (Do not infer it from any current-employment change — that would be speculation.)

---

Original assignee

Facebook, Inc. (Delaware corporation), 1601 Willow Road, Menlo Park, CA 94025.

• Primary line of business: social networking platform (Facebook / news feed / messaging).
• Did it ship a product embodying the claims? Yes. The patent claims a spam/asset-classification pipeline built on time-windowed event counters (URLs shared, hidden, clicked; unique-user vs. total counts feeding a decision-tree/model score). This maps directly onto Facebook's production news-feed integrity and anti-spam infrastructure — the specification itself uses Facebook's own URL-sharing example (http://www.xyz.com) as the worked case.
• Current status: Operating, renamed — not acquired, not dissolved, not in bankruptcy. Facebook, Inc. filed an Amended & Restated Certificate of Incorporation in Delaware on 2021-10-28 changing its name to Meta Platforms, Inc. (confirmed by the company's own "Notice of Name Change" filings in multiple federal actions, e.g. N.D. Cal. 3:19-cv-07071). Google Patents carries the name change on this patent as "2022-01-27 Assigned to META PLATFORMS, INC. … CHANGE OF NAME … Assignors: FACEBOOK, INC."

---

Assignment timeline

Two instruments are reflected on the patent's legal-events record. Reel/frame fields are unverified (see sourcing note).

1. Executed on/about 2012-09-28 (filing) / recorded 2013-03-09 — Reel/Frame [UNVERIFIED]

• Conveyance: Assignment of assignors' interest
• Assignor: Bhalchandra Suresh Agashe; Gil Shklarski; Christopher Alexander Stein; Nickolay Vladimirov Tchervenski (individually)
• Assignee: Facebook, Inc.
• Correspondent: not retrievable in this session — no recurrence determination possible.
• Context: Ordinary employee-invention assignment to the operating company at/around filing; establishes Facebook as sole owner.

2. Executed 2021-10-28 / recorded 2022-01-27 (per Google Patents legal events) — Reel/Frame [UNVERIFIED]

• Conveyance: Change of Name
• Assignor: Facebook, Inc.
• Assignee: Meta Platforms, Inc. (Delaware corporation)
• Correspondent: not retrievable for this patent. (For orientation only: the trademark counterpart of this same name change was recorded 2022-04-01 at Reel 7678/Frame 0355 with correspondent Melissa Capotosto, Kilpatrick Townsend & Stockton LLP, 1100 Peachtree St NE, Suite 2800, Atlanta, GA 30309. That is a trademark record — cite it as context, not as this patent's reel/frame.)
• Context: Pure corporate change of name only — no change in beneficial ownership; same legal entity, same assets, same address history (1601 Willow Rd → 1 Meta Way).

No other assignments appear in the record. There is no transfer to any "IP / Licensing / Holdings / Ventures" entity, no security agreement, no merger, no release, and no post-issuance third-party acquisition.

---

Timeline diagram

timeline
    title Ownership of US 9031883
    2012 : Filed by Facebook Inc
    2013 : Inventors assign interest to Facebook
    2015 : Patent 9031883 issued
    2021 : Facebook renamed Meta Platforms Inc
    2022 : Change of name recorded at USPTO

---

NPE / troll-pattern signals

---

Verdict

Operating-company assertion.*

The only recorded ownership events are (1) the inventors' 2013-03-09 assignment of interest to Facebook, Inc. and (2) Facebook's change of name to Meta Platforms, Inc. (executed 2021-10-28, recorded 2022-01-27). No NPE signal is present: there is no shell-entity transferee, no known asserter, no cascading transfer, no bankruptcy sale, and no defensive-aggregator terminator. The patent remains with its original operating-company owner under a new corporate name.

* Caveat on the label. Of the five permitted verdicts, "Operating-company assertion" is the only one that describes an operating company, and it fits the ownership half cleanly — Meta ships products embodying these claims. The assertion half is not evidenced: I found no infringement suit naming this patent, so this is more precisely an operating-company-owned, non-asserting patent. I did not select "NPE — high/moderate" (no supporting signals) or "Defensive/non-asserting" (chain does not end at a defensive aggregator). If your taxonomy permits, treat this as non-NPE, operating-company ownership retained.

Verify at: USPTO Assignment Center — https://assignmentcenter.uspto.gov/ (search patent 9,031,883 / application 13/631,523); mirror index at https://assignment.uspto.gov/patent/index.html. Outstanding data gap: pull the two reel/frame numbers and the recorded correspondents directly from Assignment Center to close the one "unclear" signal above.

I'll pull the front-page references and cited prior art for US 9,031,883 from USPTO/Google Patents sources.

Let me get the actual "Patent Citations" and "References Cited" tables for US 9,031,883.

Let me retrieve the patent's front-page PDF, which contains the authoritative (56) References Cited list.

I hit the search-step limit before I could pull the authoritative front-page (56) table from the '883 patent's own PDF, so I'll be explicit about what is verified and what is not. No fabrication.

---

0. Target confirmation (searching "9031883" literally)

The number 9031883 / US 9,031,883 B2 resolves to exactly one document:

• US 9,031,883 B2, "Systems and methods for event tracking using time-windowed counters", App. No. 13/631,523, filed 2012-09-28, granted 2015-05-12; inventors Agashe, Shklarski, Stein, Tchervenski; assignee Facebook, Inc. (now Meta Platforms, Inc.) — https://patents.google.com/patent/[US9031883B2](/patent/US9031883B2)/en and https://uspto.report/patent/grant/9,031,883

Two look-alike traps I want to flag rather than silently "correct":

1. A different patent designated "the '883 Patent" surfaced in a Sonos-related PTAB/IPR exhibit (IPR2019-00305 re U.S. Pat. 9,294,799). That document shows a face-of-patent reference list of audio/network art — Namatame 7,308,188; Fitzgerald 7,310,334; Tsuk 7,312,785; Meenan 7,313,384; Pulito 7,313,593 — and refers to a terminal disclaimer filed on "the '883 Patent application." Those references are inconsistent with Facebook's social-networking '883. I therefore treat that hit as a likely different patent / false positive and did not import it into the analysis below. (Source: docketalarm PTAB exhibit snippet.) If the reviewer intends that document, it should be checked against its own cover page before use.
2. US 10,776,437 B2 (Gupta et al., "Time-Window Counters for Search Results on Online Social Networks", Facebook) is similarly titled but has different inventors/claims — not the '883 family.

---

1. Method, sources, and one important caveat

• Sources queried: Google Patents, USPTO.report, Justia Patents, FreePatentsOnline, Patents-Review, PatentsEncyclopedia, patentimages (PDF), plus a PTAB exhibit hit.
• What I could NOT retrieve in this session: the verbatim (56) References Cited block printed on the face of the '883 patent itself. Google Patents rendered the specification but not the citation table, and I ran out of search steps before reaching the front-page PDF.
• What I DID recover: the Referenced Cited list printed on the face of a family member — the continuation US 11,636,166 (and, in part, US 9,558,279). Family members carry over most of the parent application's citations, so this is a strong but not identical proxy for the '883 list. Treat the citations below as "family-carried, pending confirmation against the '883 cover page."

Source for the citation list: https://patents.justia.com/patent/[11636166](/patent/11636166) (Referenced Cited) and https://patents.justia.com/patent/[9558279](/patent/9558279).

Contradiction flagged vs. the earlier section: the previously generated summary correctly noted the claim text was unavailable. I have since located the pre-grant publication claims (US 2014/0095412 A1, claims 1–20) — https://www.patentsencyclopedia.com/app/20140095412. These are the as-filed claims; the issued '883 claims may differ by amendment and must be confirmed. As filed, claim 1 recites: "generating … at least one time series of occurrences of an event type associated with at least one asset; determining … a first signal value and a second signal value based on the at least one time series; and classifying … the at least one asset based on comparison of the first signal value and the second signal value."

---

2. Cited references — citations and §102 mapping

Because I could not verify the reference texts in this session, I distinguish (i) bibliographic facts I could confirm from (ii) characterizations I could not verify, and I give the §102 mapping as a conditional framework rather than an assertion of anticipation. Under §102 a single reference must disclose every element of a claim, arranged as claimed — a high bar for the broad-but-specific claim 1 combination (time series + two signal values + classification by comparing them).

(A) U.S. patent documents carried on the family face

Anticipation read: On the as-filed claim 1, A2 (Jacoby) is the only one that plausibly touches the time-series/event-tracking core; A3/A4 are the only ones that plausibly touch classification. None of A1–A4, as I can characterize them, appears to disclose all of claim 1's elements — specifically the derivation of two distinct signal values from one time series and classification by comparing them. So I would not assert §102 anticipation; these read as §103 combinations (tracking reference + classification reference).

(B) Citations that are family members / post-dating documents (NOT '883 prior art)

These appear on the family list but cannot anticipate the '883 (same inventors or later filing):

• US 2014/0095412 A1 — Agashe et al., pub. 2014-04-03 — this is the '883's own pre-grant publication.
• US 9,031,883 B2 — Agashe et al., issued 2015-05-12 — the target itself, cited by the continuations.
• US 2015/0199444 A1 — Agashe et al., pub. 2015-07-16 — continuation (US 9,558,279) publication.
• US 2014/0244644 A1 — Mashinchi et al., pub. 2014-08-28 — post-'883; continuation-era citation.

(C) Foreign patent documents and non-patent literature (NPL)

I did not retrieve any foreign-patent or NPL citations for the '883. I cannot state whether any exist. Unverified.

---

3. Bottom line

• Patent identified literally as US 9,031,883 B2 (Facebook / Meta), no auto-correction applied; the similarly-numbered "883" in a Sonos PTAB exhibit is flagged as a likely different patent and excluded.
• Most relevant cited art (for §102 purposes): none of the references I could recover is a clean single-reference anticipation of as-filed claim 1. The references cluster into (i) event/time tracking (Jacoby 2012/0041711 — closest on the counter/time-window limitations) and (ii) classification/flags (Byrd 2012/0096004, Rodriguez 2012/0173710 — closest on the spam/popularity/ranking limitations). Mahmoud 8,704,662 postdates the filing and is not '883 art on its face; Fischell 2009/0082682 is the earliest but its content is unverified.
• Claims potentially implicated if a reference is confirmed to disclose the corresponding structure: 1 (all independent aspects); 3–5 (time window/intervals/counters); 8 (circular buffer reset); 9–12 (linear buffer, decay, counter roll-over); 13–15 (score/decision-tree/ML; spam/popularity/ranking); 16 (integral); 17 (policy modification); 18 (real-time); 19–20 (system/CRM). Anticipation by any single one of A1–A4 is not supportable on the record I have.

4. Caveats (read before relying)

1. The '883's own (56) table was not retrieved. The list above is from family-member face pages. A reviewer with USPTO PatentCenter/IFW access should pull the '883 grant cover page to confirm the exact set, including any foreign documents and NPL the examiner cited.
2. Reference content was not read in this session, so the §102 mappings are conditional, not asserted anticipations.
3. As-filed ≠ issued claims. The claim numbers used above are from the 2014/0095412 pre-grant publication; confirm against the issued '883 claims.
4. If the task intends the Sonos-related '883 rather than Facebook's, that is a different document and requires a fresh search keyed to its own number.

I need to gather the actual prior art before analyzing. Let me note upfront: the "Prior Art section of this page" was not included in what I received — I have the specification, the previously generated summary/litigation sections, but no prior-art reference list. Let me try to retrieve the prior art cited on the patent and related references.

Let me try to retrieve the actual front-page "Patent Citations" list and any prior-art references tied to this patent's family.

Obviousness Analysis — U.S. Patent No. 9,031,883 B2

("Systems and methods for event tracking using time-windowed counters") under 35 U.S.C. § 103

---

0. Threshold caveat — the "Prior Art section" referenced in the task was not provided to me

Before the merits: I must flag a source gap up front, per my operating rules.

• The task instructs me to "use the results from the Prior Art section of this page." No such section was included in the material provided to me. What I received was (i) the full specification text of '883 (stopping before the claims), and (ii) the previously generated patent-summary and litigation-summary sections. There is no appended prior-art reference list, and my searches did not surface the verbatim front-page "Patent Citations" table for US 9,031,883 (the Google Patents page returns the AI-generated "Definitions" block rather than the citation tables; the PDF citation table was likewise not retrieved).
• Consequently, I cannot represent that the references below are "the Prior Art section" results. I will not fabricate that provenance.
• What follows is a § 103 framework built on (a) the claim scope reconstructed from the patent's own SUMMARY (as in the prior section), and (b) well-known, real, pre-2012 references and techniques that a person of ordinary skill in the art (POSITA) would have had constructive knowledge of — each explicitly labeled as candidate/representative and requiring verification against the actual prosecution record and the page's Prior Art section. If you paste that Prior Art section, I will re-map it element-by-element and replace the candidates with the record references.

Contradiction/flag: the earlier "Segments" note stated the claim text was unavailable and reconstructed the claims from the SUMMARY. I inherit that reconstruction and flag it again — the mapping below is only as reliable as that reconstruction and must be checked against the printed claims.

Two secondary items that surfaced in my searches and are not prior art to '883, listed so they are not mistaken for it:

• US 10,776,437 B2 (Gupta et al., "Time-Window Counters for Search Results on Online Social Networks," Facebook, filed 2017-09-12) — later than '883; a family-adjacent/technical cousin only, not § 102 art.
• US 2013/0303204 A1 (Sybase 365, "Method of and system for detecting spam activity in a cloud system") — publication date after the '883 priority date, but if its US filing date is indeed 2012-05-08 (as a search snippet indicated), it is a potential pre-AIA § 102(e) reference; flag for verification.

---

1. Governing legal standard (pre-AIA § 103(a))

Because '883 was filed 2012-09-28 — before the AIA's 2013-03-16 effective date — pre-AIA § 103(a) governs, and "prior art" is defined by pre-AIA § 102(a)/(b)/(e)/(g).

Under Graham v. John Deere Co., 383 U.S. 1 (1966), and KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007):

1. Scope and content of the prior art.
2. Differences between the prior art and the claims.
3. Level of ordinary skill at the time of the invention (here, ~2012).
4. Secondary considerations (objective indicia: unexpected results, long-felt need, commercial success, industry skepticism, copying).

KSR is decisive framing. Where a claim is "the product of ordinary creativity, not innovation" — i.e., a predictable combination of known elements to achieve a known purpose — it is obvious. Motivation to combine need not be found expressly in the references; it may flow from the problem to be solved, the known desirability of the combination, and the fact that the field presents "a finite number of identified, predictable solutions." KSR, 550 U.S. at 421. Also, design incentives / market forces and "obvious to try" reasoning apply. I will return to the teaching-away counter-argument at § 7 — notably, the specification itself almost hands it to a challenger.

---

2. The claimed subject matter, broken into elements (reconstructed)

The alleged inventive core (per the prior section): the combination of (a) bucketed/windowed counters in circular or decaying/fixed linear buffers; (b) derivation of plural signal values from the resulting time series — optionally along a uniqueness dimension; and (c) automated classification by comparing those signals — aimed at spam/popularity/ranking. Critically, the point of novelty is the use of the counters, not the counters themselves, and not any novel buffer, counter, statistic, or classifier. Every individual building block was independently known; the question is whether their assembly was nonobvious.

---

3. Level of ordinary skill (POSITA, ~2012)

A POSITA would be a software/systems engineer or data scientist with a bachelor's degree in CS/EE (or equivalent) and ~2–4 years' experience in one or more of: large-scale distributed systems/data infrastructure, streaming/online analytics, social-network or web anti-abuse, and applied machine learning / statistics. That person would know, without undue experimentation: ring/circular buffers; bucketed sliding-window counters; time-decay aggregation; distinct-count estimation over streams; spam-filtering classifiers; and the short-window/long-window "ratio" pattern for anomaly detection. This skill level matters because nearly every element recites a routine engineering technique.

---

4. Prior-art landscape — candidate/representative references (verify against the record)

Verification status: the following are real, pre-2012 references or techniques I hold with high confidence; they are offered as representative of the state of the art, pending the actual Prior Art section. They are not confirmed to be the examiner-cited references.

Group A — Windowing, circular buffers, and time-decay counters (elements 1B, 1C, 3, 4, 5)

• Knuth, The Art of Computer Programming, Vol. 1 — circular/ring buffer as a fundamental data structure; overwrite-oldest semantics. (Element 3.) Ancient, ubiquitous.
• Varghese & Lauck, "Hashed and Hierarchical Timing Wheels," SOSP 1987 — a timer facility using multiple counter levels of progressively coarser granularity, with values carried forward from a finer level to a coarser level as time advances. This is a direct structural analogue of the '883 linear buffer with exponentially-decaying intervals and the "value provided from a first counter to a second counter, first counter reset" limitation (element 4). (High confidence — this is a landmark data-structures paper.)
• Cormode, Shkapenyuk, Srivastava & Xu, "Forward Decay: A Practical Time Decay Model for Streaming Systems," ICDE 2009 — formal time-decay aggregation over data streams; sliding-window and exponentially-weighted aggregations, including how to combine multiple decay/window scales. (High confidence. Directly on point for "signal values" from decays over windows.)
• Flajolet et al., "HyperLogLog: the analysis of a near-optimal cardinality estimation algorithm," 2007 (and its predecessor Flajolet–Martin 1985) — distinct-count / uniqueness estimation over data streams. (Element 2.) Extremely well known by 2012.
• Textbook EWMA/short-vs-long-running-average techniques (also standard in network traffic engineering) — comparing aggregates over two different window lengths.

Group B — Spam detection in social networks and time-series abuse signatures (elements 1A, 7, 9, 10)

• Benevenuto, Magno, Rodrigues & Almeida, "Detecting Spammers on Twitter," CEAS 2010 — classifies accounts as spammers using feature vectors derived from rates and ratios of user actions (e.g., fraction of tweets that are replies/URLs, action frequency). (High confidence. Establishes that abuse detection via action-rate features was conventional by 2010–2012.)
• Gao, Hu, Li, Zhang, Wilson & Chen, "Detecting and Characterizing Social Spam Campaigns," IMC 2010 — detects spam campaigns by patterns of URL-sharing/click activity over time, including rapid spreading of a URL by many accounts. (High confidence.)
• General anti-abuse practice on large platforms (2010–2012): track per-asset actions (share/click/hide) over rolling windows; apply thresholds/rules. (Well-known industry practice; supports elements 1A, 7.)

Group C — Classifiers over derived features (elements 6, 7, 8)

• Sahami, Dumais, Heckerman & Horvitz, "A Bayesian Approach to Filtering Junk E-Mail," AAAI Workshop 1998 — ML/Bayesian classification of spam from features; the canonical reference that automated spam classification from numeric features is old and routine. (High confidence.)
• Standard decision-tree / ML classifiers (Quinlan's ID3/C4.5, decision-tree literature; logistic regression; boosted trees) — textbook by 2012. (Element 6.)
• The integral/area-under-curve (element 8) is elementary calculus and standard for aggregating a rate/series over an interval.

Group D — The "compare two window aggregates" classification hook (element 1D/1E)

• The short-term vs. long-term statistic comparison is the classic anomaly/change-detection pattern (e.g., ratio of a short-window average to a long-window average; EWMA crossover). This is exactly the "compare first signal value and second signal value" limitation, and it was pedestrian in network monitoring and anomaly detection long before 2012.

Net: Each element — buffered interval counters, decay/carry-forward, uniqueness, multi-window aggregates, ML classification of spam, and the short/long comparison — maps to an independent, long-known reference. The claim is their aggregation, which is the paradigmatic KSR fact pattern. (No single reference is asserted here as an anticipatory § 102 reference; the analysis is combination-based § 103.)

---

5. Representative combinations rendering the claims obvious

Combination 1 (primary) — A (windowed counters) + B (spam-rate features) + C (ML classifier), optionally + D (short/long comparison)

What each contributes:

Motivation to combine (KSR):

1. Same field and same problem. Both A and B address monitoring user activity in real time, and B specifically addresses social-network abuse. Combining a known streaming-aggregation substrate with a known abuse-detection application is not a leap.
2. The problem itself supplies the motivation. The operator's stated need — distinguish 100 shares by one user (spam) from 100 shares by 100 users (popularity) — is precisely the uniqueness problem that Group A's distinct-counting solves, and precisely the discrimination that Group B's rate/ratio features make. A POSITA seeking to catch spam in a URL-sharing product would naturally (i) count shares per asset in rolling windows and (ii) add a unique-shares count to separate spam from popularity — the specification says exactly this.
3. Predictable result. Feeding windowed counters into a classifier yields the predictable, expected benefit (earlier/better spam detection). No new technical effect is asserted.
4. "Finite number of predictable solutions." Given rolling per-asset counters, there are only a handful of obvious ways to organize them (ring buffer, fixed buckets, decay/carry-forward) and only a handful of obvious ways to classify from them (thresholds, decision trees, ML). KSR 550 U.S. at 421.

Combination 2 — A + D alone (no ML): windowed counters + short/long-window ratio with a threshold rule

Even setting Group C aside, a POSITA would find obvious the specific "compare first signal value to second signal value" step, because comparing a short-window rate (f₁ = last hour) against a longer-window rate (f₂ = last 3 hours) to flag a spike/anomaly is among the most basic monitoring heuristics. The patent's own worked example (f₁, f₂, g₁, g₂) is a textbook multi-window comparison. This combination disposes of independent claim 1 and dependent claims 8, 10.

Combination 3 — A + B: windowed uniqueness counters to distinguish spam from popularity

Track, per asset, two counters: total shares and unique-user shares, over a rolling window; when total ≫ unique (concentration in few users), flag as suspect. Group A supplies the uniqueness/streaming-count machinery (element 2); Group B supplies the spam-vs-popularity distinction (element 7). This disposes of claim 1 plus dependent claims 2 and 7.

Motivation — KSR's "design incentive" prong: each combination targets the acknowledged problem (spam vs. popularity in a sharing product) using off-the-shelf tools, so the motivation is supplied by the problem and by the known desirability of the improvement.

---

6. Dependent-claim snapshots

---

7. The strongest rebuttal the patent owner can make — and why the specification undercuts it

A skilled owner/PO would argue nonobviousness from the specification's own "teaching-away"/design-choice passages, namely the discussion that:

• circular buffers lose the ability to compare like periods across time (overwrite);
• exponentially-decaying buffers lose granularity for older events;
• non-decreasing buffers cost more storage but preserve precision for period-over-period comparison.

Why this is weak under KSR: the specification frames these as three known implementations of one idea, with known, predictable tradeoffs ("[t]he selective use of certain buffers or time intervals accordingly may pose advantages"). Presenting a menu of known alternatives with predictable costs/benefits is the definition of an obvious design choice — not a teaching-away. A teaching-away requires that the reference criticize, discredit, or discourage the claimed combination; here the spec endorses each variant for a suitable use. If anything, the text is an admission of the obvious design space.

The owner might also argue unexpected results — but the spec asserts no data showing a surprising synergy; the stated benefit (distinguish spam from popularity) is the expected result of adding a uniqueness dimension.

---

8. Secondary considerations / objective indicia

• No evidence in the record I have (and none surfaced) of unexpected results, long-felt-but-unmet need, industry skepticism, copying, or nexus-based commercial success attributable to the claimed combination. The apparent "Approx Value" attribution seen on a third-party analytics site (patentleaderboard) is a heuristic valuation, not evidence of commercial success or nexus — do not treat it as an objective indicium.
• Patent-family context: the existence of continuations with the same title (US 9,558,279; US 10,740,419; US 11,636,166) is a portfolio/breadth fact, not a nonobviousness argument. (Consistent with the prior family note; no contradiction.)

---

9. Cross-reference / adjacent issue (flag, not the subject of this task)

The claims also invite an Alice § 101 challenge (classifying assets by comparing aggregate numbers is arguably an abstract idea implemented with generic computing). That is outside the present § 103 task, but it is worth noting that the same facts that make the combination obvious under KSR (routine counters + routine ML + routine comparison) also support the "well-understood, routine, conventional" step-two inquiry. I flag it; I do not analyze it here. There is no contradiction with the prior sections — the prior litigation section found no assertion of '883, and the prior claim overview matches the reconstruction used above.

---

10. What I could not verify, and the recommended next step

1. I did not receive the "Prior Art section," and my searches did not return the verbatim front-page citation list for US 9,031,883. Therefore the references in § 4 are candidate/representative state-of-the-art references, not confirmed record prior art.
2. The claim text remains reconstructed from the SUMMARY (per the prior section). Element-by-element precision depends on obtaining the printed claims.
3. The § 102(e) status of US 2013/0303204 A1 depends on its actual US filing date (a snippet suggested 2012-05-08) — verify.

To finalize a filing-grade obviousness position, provide:

• the Prior Art section / front-page "Patent Citations" for '883 (and, if relevant, the sibling '279/'419/'166 lists), and
• the printed claim set.

On receipt I will (a) substitute the record references for the candidates above, (b) re-run the element-by-element mapping in § 5 against the actual limitation language, and (c) add an express motivation-to-combine statement keyed to each reference's own disclosure, in the form used in an IPR petition or office action.

Bottom line (based on available evidence): on the claim scope as reconstructed, the '883 claims appear obvious under pre-AIA § 103(a), because they recite the predictable assembly of individually known elements — windowed/decaying counters (Group A), unique-user counting (Group A), ML/decision-tree spam classification (Group B/C), and a short-vs-long-window comparison (Group D) — each directed to the well-known problem of distinguishing spam from genuine popularity in a social network, with the motivation supplied by the problem itself and the combinations falling within the "finite number of identified, predictable solutions" recognized in KSR. This conclusion is subject to the source gap flagged in § 0 and should be confirmed against the actual Prior Art section and printed claims before reliance.