https://eureka.patsnap.com/resources/blog/rd-blog/data-poisoning-and-backdoor-defense-patent-landscape/ · Patsnap · data cut-off 2026-07-31 · downloaded from the live page
Patent Landscape · Data Curation & Governance
Data Poisoning and Backdoor Defense Patents: Who Holds the Claim Space
  • 50.0% concentration. The top 5 assignees hold 26 of the 52 records in scope — half the field sits with a small group of filers.
  • Filing fell 86% from 2021 to 2024. That is the last complete year in this dataset; 2025 and 2026 figures will still rise as publication catches up.
  • Agricultural IPC classes dominate. A01M and A01D together cover pest control and harvesting equipment, an unusual technical center of gravity for a defensive-AI topic.
Get a prior-art report on your approach
52
Published Records
50%
Top-5 Share of All Records
-86%
Filing Growth 2021→2024
US
Leading Jurisdiction

Filing growth compares 2021 (7 records) with 2024 (1) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 52 records in scope (CR5), not by the ranked leaders only.

Published byPatsnap Research··7 min readSourced from Patsnap Eureka
Field Overview

What this landscape covers

This review covers 52 published records matched against poisoning and backdoor terminology combined with defense-oriented terms such as clean label attack, poison rate, spectral signature and certified defense. The dataset spans filings from 2015 through the 2026-07-31 cut-off, with the earliest activity in this search appearing from 2017 onward.

The technology composition is split unevenly between machine-learning security terms (G06N, H04L, G06F, G06V) and physical agricultural equipment classes (A01M, A01D, A01B). That split matters for anyone searching this space: a query built only on AI-security keywords will miss a meaningful share of the prior art sitting in equipment and material-analysis classes.

Filing activity and IPC composition, 2017–2026
  1. 1DEERE & CO10
  2. 2SCINTILLON INST FOR BIOMEDICAL & BIOENERGY RES5
  3. 3NORTH DAKOTA STATE UNIV RES FOUND4
  4. 4HEWLETT PACKARD ENTERPRISE DEV LP4
  5. 5NOKIA SOLUTIONS & NETWORKS OY3
  6. 6SURROMED INC3
  7. 7BARREL PROOF TECHNOLOGIES LLC3
  8. 8WIZ INC2
  9. 9STRONG FORCE EE PORTFOLIO 2022 LLC2
  10. 10STRONG FORCE IOT PORTFOLIO 2016 LLC2
Source: Patsnap Eureka. Assignee ranking and totals. Derived from a Patsnap search on Data Poisoning and Backdoor Defense covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP

Let an AI agent run this analysis on your own technology

Pick a task. Every answer cites the patents behind it.

10,000 free credits to start
Filing Data

Trend and technology composition

The two views below are the record of what has actually been filed and published in scope — not a forecast. Read the most recent one or two years as a floor, not a ceiling, given the usual 18-month gap between filing and publication.

Filing trend, 2017–2026

Filings rose from zero in 2017 to a peak of 9 in 2025, but the only clean year-over-year comparison the data supports is 2021 (7 filings) against 2024 (1 filing) — an 86% drop across that span. Treat 2025 and 2026 counts as still filling in.

Filing trend, 2017–20260358100201720182019202020212022202320249202572026Most recent year is partial — publication lag means later filings are not yet visible.

IPC subclass composition

G01N (material analysis & testing) and G06N (AI-model computing) are tied at 23.1% of the 52 records each, followed closely by A01M (pest & vermin control) and H04L (digital transmission) at 19.2% each. Because records can carry multiple IPC codes, these percentages are each measured against the full 52-record total and will not sum to 100%.

IPC subclass compositionG01N · Material analysis & testing1223.1%G06N · Computing based on AI models1223.1%A01M · Pest & vermin control1019.2%H04L · Digital information transmissi…1019.2%A01D · Harvesting & mowing917.3%G06F · Electric digital data processi…815.4%G06V · Image/video recognition815.4%A01B · Soil working in agriculture611.5%Other3261.5%

Shares are the percentage of the 52 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.

Source: Patsnap Eureka. Filing trend and technology composition. Derived from a Patsnap search on Data Poisoning and Backdoor Defense covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.

Go deeper on Data Poisoning and Backdoor Defense with Eureka

This page is one run against one query. Ask Eureka your own question about data poisoning and backdoor defense and every answer comes back with the patent numbers behind it.

Try Eureka
Key Patents

Representative filing and most-cited records

Representative Record
US9867012B22018-01-09

US9867012B2 — Whispered speech detection (DSP Group)

DSP GROUP LTD.

A method for detecting whispered speech of a user of a mobile computerized device, comprising detecting whispered speech context and attempting to detect the whispered speech using detection parameters that differ from those used for non-whispered speech.Published 2018-01-09. Included in scope as a representative record; its detection-parameter approach is adjacent to trigger-detection claim language used elsewhere in this field.

US9867012B2 — patent drawing 1US9867012B2 — patent drawing 2
View full record
Most-cited records in scope
#Publication no.Patent titleCitations
1US6625399B1Non-linear flash and lightning detection device79
2WO2012103138A1Automatic detection of appliances43
3WO2002040698A2Analysis of soluble factor biological markers35
4WO2025160415A1Systems, methods, devices, and platforms for industrial internet of things31
5US20210243950A1Agricultural harvesting machine with pre-emergence weed detection and mitigation system30
6US20010050339A1On-the-go sugar sensor for determining sugar content during harvesting24
7US20210243938A1Pre-emergence weed detection and mitigation system22
8US6624888B2On-the-go sugar sensor for determining sugar content during harvesting18
9US20040021862A1On-the-go sugar sensor for determining sugar content during harvesting16
10US6851662B2On-the-go sugar sensor for determining sugar content during harvesting16

Citation counts reward older filings that have had more time to accumulate citations within this corpus; they are a signal of influence, not of current technical importance.

Each row carries its publication number; clicking a row searches Eureka by that number.

Source: Patsnap Eureka. Citation counts and representative records. Derived from a Patsnap search on Data Poisoning and Backdoor Defense covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Run it yourself

Put your own technology through the same analysis


  
Where to run it
Fastest

Eureka on the web

When you want the answer in the next five minutes.

The agent works the prompt against patents and technical literature, citing every source.

Run your analysis now →
For builders

MCP server & REST API

When it has to run inside your own pipeline.

Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.

Browse MCP servers →
Insights

What the numbers mean for filing strategy

Three patterns stand out once you look past raw counts: where filings concentrate, where they don't, and what the trend line can and cannot tell you.

Concentration
50.0% / 26 of 52
top 5 combined share

Half the field sits with five filers

The top 5 assignees combined hold 26 of the 52 records in scope, exactly half. The leader alone accounts for 10 records, with a sharp drop to 3 records at fifth place — a steep-shouldered distribution rather than a gradual taper.

Concentration figures are against the full 52-record total, not the ranked list.
Long tail
24 ranked assignees
companies in the ranking

A long tail below the top 10

The ranking returns 24 companies; the top 10 combined hold 73.1% of all 52 records, meaning the remaining 14 ranked assignees split the rest across single- or double-digit filings each.

This is the full ranking the data endpoint returns, not a top-50 or top-100 cut.
Momentum
-86% (2021→2024)
last complete comparison

A sharp recent-year drop, with caveats

Filings fell from 7 in 2021 to 1 in 2024, an 86% decline across that span. One assignee shows 3 filings in the latest year in this dataset, suggesting renewed activity that predates full publication of the broader field.

2025 and 2026 counts will rise as publication catches up; do not read them as a slowdown.
Technology mix
23.1% G01N & G06N
share of 52 records

Material testing and AI computing are tied at the top

G01N (material analysis & testing) and G06N (computing based on AI models) each cover 23.1% of the 52 records, with agricultural classes A01M and A01D close behind — an indication that trigger-detection and poisoning-defense claims are being filed as much through physical inspection and equipment routes as through pure ML-security language.

Class shares sum to more than 100% because records carry multiple IPC codes.
Eureka AI Agent
Looking for what nobody has claimed yet?

Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to data poisoning and backdoor defense, with the prior art for and against each one.

Find the white space →
Source: Patsnap Eureka. Co-assignee relationships and derived observations. Derived from a Patsnap search on Data Poisoning and Backdoor Defense covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Players & Positioning

Who holds the claim space, and where it opens up

The leadership group is compact, but the gap between the top 10 and the rest is where most of the remaining opportunity to file sits.

Leader
10 records
leading assignee

A single filer sets the pace

The leading assignee holds 10 of the 52 records in scope, well ahead of the fifth-place holder at 3. That gap suggests the leader has built out a broad claim footprint rather than a narrow, defensible position.

Leadership by record count, not by family-adjusted count.
Fresh entrant
3 in latest year
recent-year momentum

New activity from outside the historical leadership

Barrel Proof Technologies shows 3 filings in the latest year, more recent activity than several assignees with larger historical totals — worth tracking even though single-year counts from the most recent period are still incomplete.

Momentum figures reflect publications received so far, not final totals.
Established but quiet
0 in latest year, -100% YoY
recent-year momentum

Historical filers have gone quiet

Hewlett Packard Enterprise shows a -100% year-over-year change with zero filings in the latest year, alongside other established assignees with zero recent activity — a pattern consistent with the broader post-2021 filing pullback in this dataset.

A pullback in publications does not necessarily mean a pullback in filing, given the 18-month lag.
🔍
Under-claimed sub-areas worth a closer look
These branches show thinner filing density relative to the core AI-security classes, based on the IPC composition in scope.
Spectral-signature poison detectionClean-label attack certification methodsCross-domain trigger detection (vision-to-agriculture)Certified defense bounds for poisoned dataData supply-chain provenance verification
Rank all filers by momentum →
Recent-year filing momentum by assignee
AssigneeRecent yearYoY
BARREL PROOF TECHNOLOGIES LLC3—
Deere & Co.10%
SCINTILLON INST FOR BIOMEDICAL & BIOENERGY RES0—
Hewlett Packard Enterprise Development LP0-100%
Nokia Solutions and Networks Oy0—
SURROMED INC0—
Tencent Technology (Shenzhen) Co., Ltd.0—
Wiz Inc.0—
Source: Patsnap Eureka. Assignee-level momentum. Derived from a Patsnap search on Data Poisoning and Backdoor Defense covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
What's Next

Where to take this analysis

The dataset points to a concentrated top and a genuinely open middle. The next steps depend on whether you are clearing a filing or tracking a competitor.

Clear a filing against the leadership group

Run a freedom-to-operate check against the top 5 assignees before drafting claims in G01N or G06N — that is where half the field's records sit.

Explore assignee claims in Eureka

Track momentum shifts as 2025–2026 data fills in

Recent-year counts will rise as publication catches up; set a watch on assignees showing early activity rather than relying on the 2024 snapshot alone.

Set up monitoring in Eureka

Test claim language in the under-claimed branches

Spectral-signature detection and clean-label certification show thinner density than the core AI-security classes — a first-filer position may still be available there.

Draft and search claims in Eureka
Source: Patsnap Eureka. Forward-looking reading of the same dataset. Derived from a Patsnap search on Data Poisoning and Backdoor Defense covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
FAQ

Common questions about this landscape

Answers are grounded in the same dataset. Derived from a Patsnap search on Data Poisoning and Backdoor Defense covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP

Research Data Poisoning and Backdoor Defense in depth with Eureka

Go past this page: query the whole data poisoning and backdoor defense corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.

Try Eureka

Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.

Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.

Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.

Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company's registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.

Help us improve

Something wrong, out of date, or missing on this page? Tell us and it goes straight to the team that maintains it.