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 →https://www.patsnap.com/resources/blog/rd-blog/machine-learning-for-vulnerability-detection-patent-landscape/ · Patsnap · data cut-off 2026-07-31 · downloaded from the live page
Patent Landscape · 2,271 recordsMachine Learning for Vulnerability Detection: Patents Mapped Across Filers, Classes and Time
- Filings peaked in 2021 at 339 and have since eased to 238 by 2024, a -30% move over that span — publication lag means 2025-2026 figures will still fill in.
- No single filer dominates the ranked leader holds 56 records and the top 5 combined account for just 11.1% of all 2,271 records in scope.
- Two classes anchor the field G06F and G06N each cover roughly 47% of records, while healthcare informatics (G16H) and bioinformatics (G16B) show unexpected overlap with vulnerability-detection filings.
2,271
Published Records
11%
Top-5 Share of All Records
-30%
Filing Growth 2021→2024
US
Leading Jurisdiction
Filing growth compares 2021 (339 records) with 2024 (238) — 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 2,271 records in scope (CR5), not by the ranked leaders only.
Published byPatsnap Research··7 min readSourced from Patsnap Eureka
Overview
What this landscape covers
This landscape tracks patent activity where vulnerability detection and code embedding techniques intersect with machine learning concerns like label noise, cross-project transfer, and abstract syntax tree representations. The scope spans 2,271 published records filed between 2015 and mid-2026, drawn from a search string that pairs detection methods with the training and generalization problems that make them work — or fail — outside a single codebase.
The dataset captures both dedicated security vendors and general-purpose computing and AI filers whose portfolios touch code analysis incidentally. That mix matters: a claim written broadly around neural classifiers for code can read on vulnerability detection even when the filer's core business is elsewhere, which is part of why the technology composition below spreads across classes well beyond traditional security codes.
Filing activity and technology mix, 2015-2026
Source: Patsnap Eureka. Assignee ranking and totals. Derived from a Patsnap search on Machine Learning for Vulnerability Detection 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
The Data
Filing trends and technology composition
Two views of the same 2,271 records: how filing volume has moved year over year, and which IPC subclasses the claims actually sit in.
Filing trend: a 2021 peak, then a cooling that is still being written
Annual filings rose from 81 in 2017 to a peak of 339 in 2021, then declined to 238 by 2024 — a -30% move over that three-year span. Counts for 2025 and 2026 are shown as published but will understate true filing activity, since publication typically lags filing by around 18 months.
Technology composition: general-purpose AI classes dominate
G06F (electric digital data processing) and G06N (AI computing models) each appear on close to half of all 2,271 records, confirming that most filings are built on general digital-processing and AI-model claims rather than narrow security-specific ones. Image and data recognition classes (G06V, G06K, G06T) and domain-specific classes (G16H healthcare informatics, G16B bioinformatics) appear on a meaningful minority of records, since a record can carry multiple IPC classes these shares add up to more than 100%.
Shares are the percentage of the 2,271 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 Machine Learning for Vulnerability Detection covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.
Go deeper on Machine Learning for Vulnerability Detection with Eureka
This page is one run against one query. Ask Eureka your own question about machine learning for vulnerability detection and every answer comes back with the patent numbers behind it.
Try EurekaKey Patents
Most-cited records and a representative recent filing
Representative recent filing
System and method for vulnerability localization based on deep learning (US20250077683A1)
HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY
Filed by Huazhong University of Science and Technology, this record analyzes a code file to build an abstract syntax tree, enriches it with data-flow and control-flow edges to form a semantic-flow-enhanced tree, splits that tree into sub-trees, and feeds the sub-trees into a pre-established vulnerability detection and localization model.The two-stage AST enrichment — adding flow edges before splitting into sub-trees — is the structural detail that distinguishes this approach from plain syntax-only embedding methods.
Highest-cited records in scope
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20150309813A1 | A System for analyzing applications in order to find security and quality issues | 435 |
| 2 | US8522348B2 | Matching with a large vulnerability signature ruleset for high performance network defense | 268 |
| 3 | US20200258223A1 | Determining biomarkers from histopathology slide images | 239 |
| 4 | US20170060254A1 | Multi-sensor based user interface | 238 |
| 5 | US20230206329A1 | Transaction platforms where systems include sets of other systems | 233 |
| 6 | US20180121799A1 | Training a Joint Many-Task Neural Network Model using Successive Regularization | 226 |
| 7 | US20230214925A1 | Transaction platforms where systems include sets of other systems | 205 |
| 8 | US9426177B2 | Method and apparatus for detecting security vulnerability for animation source file | 200 |
| 9 | US20130103624A1 | Method and system for estimating response to token instance of interest | 194 |
| 10 | US20200320769A1 | Method and system for predicting garment attributes using deep learning | 187 |
Citation counts favor older records simply because they have had more time to accumulate citations within the searched corpus — read them as a signal of influence, not of current relevance.
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 Machine Learning for Vulnerability Detection 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
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 say about where this field stands
Three patterns run through the filing data, the citation table and the technology mix.
Concentration
11.1% of 2,271 records
held by the top 5 assignees
Filings are dispersed, not cornered
The top 5 assignees combined account for just 11.1% of all 2,271 records in scope, and the top 10 for 18.4%. With the ranked leader at 56 records against a field of 2,271, no single filer has locked up the space — a long tail of smaller and single-filing entrants still has room to stake claims.
Based on the 100-company assignee ranking returned by the dataset.
Momentum
339 → 238
peak-year to 2024 filings
Volume cooled after a 2021 peak, not yet a decline
Filings climbed from 81 in 2017 to a peak of 339 in 2021, then eased to 238 by 2024 — a -30% move over that span. Because publication lags filing by roughly 18 months, the 2025-2026 counts in the raw trend understate real activity and should not be read as a continued slide.
2021 remains the single highest-volume year in the dataset.
Technology mix
47.1% / 47.0%
share of records on G06F / G06N
General AI-computing claims outweigh security-specific ones
G06F and G06N each sit on close to half of all 2,271 records, while more specialized classes — G16H healthcare informatics at 10.2% and G16B bioinformatics at 8.2% — appear on a smaller but non-trivial share. That overlap suggests detection models built for code are drawing on techniques and claim language shared with medical-imaging and bioinformatics classifiers.
Shares sum to more than 100% because records can carry multiple IPC classes.
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 machine learning for vulnerability detection, with the prior art for and against each one.
Source: Patsnap Eureka. Co-assignee relationships and derived observations. Derived from a Patsnap search on Machine Learning for Vulnerability Detection 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
Who is filing, and where the activity is thin
The ranked leaders span both dedicated security vendors and large general-purpose technology filers; the momentum data suggests even the most active names have slowed their pace in the latest year.
Leader
56 records
top-ranked assignee
A leader without a runaway lead
The top-ranked assignee holds 56 records against a field where fifth place sits at 43 and tenth place at 28 — a gradual taper rather than a cliff. That shape points to a field where being first to file in a sub-area still carries real weight.
Ranking covers 100 assignees, counted in patent families.
Momentum
0% to -100% YoY
latest-year change among top filers
Even active filers have gone quiet recently
Recent-year momentum among several of the more active assignees shows flat or sharply negative year-over-year change, including filers moving from prior activity to zero in the latest year. Given publication lag, this is as likely to reflect filings still working through the pipeline as an actual pullback.
Momentum reflects the most recent year of published data, which is incomplete.
Collaboration
4 co-assignee pairs
identified in the dataset
Co-filing is rare and mostly intra-group
Only four co-assignee pairs appear across the dataset, and the strongest of them links two related entities under the same corporate family rather than an arm's-length partnership. Cross-organization collaboration on filings is not yet a defining feature of this space.
Strongest pair: Tempus Labs Inc and Tempus AI Inc, three shared records.
Recent-year filing momentum by assignee
| Assignee | Recent year | YoY |
|---|---|---|
| Oracle International Corporation | 1 | 0% |
| Google LLC | 1 | 0% |
| Microsoft Technology Licensing, LLC | 0 | -100% |
| TEMPUS LABS INC | 0 | — |
| International Business Machines Corporation | 0 | — |
| STRONG FORCE TX PORTFOLIO 2018 LLC | 0 | -100% |
| Intel Corporation | 0 | — |
| Strong Force TX Portfolio 2018 LLC | 0 | — |
Source: Patsnap Eureka. Assignee-level momentum. Derived from a Patsnap search on Machine Learning for Vulnerability Detection 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 next
The landscape points to open claim space rather than a settled field — the next step is usually narrower than a full re-run of this search.
Check freedom-to-operate on a specific claim shape
Before filing on an AST-enrichment or patch-classification approach, run the specific claim language against the most-cited records in this set, since citation weight signals which older filings still shape examiner reasoning.
Explore prior art in EurekaTrack the assignees still active in the latest year
Momentum data shows most top filers cooling off; watching which ones keep filing through the publication-lag window will separate genuine pullback from reporting delay.
Set up assignee tracking in EurekaScope the under-claimed branches before competitors do
Cross-project transfer and patch-level classification show thinner filing density than core detection claims — a first-mover advantage may still be available there.
Run a white space search in EurekaSource: Patsnap Eureka. Forward-looking reading of the same dataset. Derived from a Patsnap search on Machine Learning for Vulnerability Detection 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
The ranked leader in this dataset holds 56 of the 2,271 records in scope, with the field tapering gradually rather than dropping off sharply — fifth place sits at 43 records and tenth at 28. The top 5 assignees combined account for 11.1% of all records, and the top 10 for 18.4%, which means no single company controls the space. The assignee mix includes both dedicated security vendors and large general-purpose technology and AI companies whose broader portfolios happen to touch code analysis.
Filings rose steadily from 81 in 2017 to a peak of 339 in 2021, then eased to 238 by 2024, a -30% move over that three-year span. That is a cooling from an unusually high peak rather than clear evidence of decline, and the 2025-2026 figures in any raw trend chart will understate real activity because patent publication typically lags filing by around 18 months. Treat the last one to two years of any filing trend as provisional.
The core of the field sits in general digital-processing and AI-model claims: G06F and G06N each appear on close to half of the 2,271 records in scope. Beyond that, image and data-recognition classes (G06V, G06K, G06T) and domain-specific classes like G16H healthcare informatics and G16B bioinformatics also show meaningful overlap, since detection models for code often reuse claim structures built for medical-imaging or bioinformatics classifiers. Because a single record can carry several IPC classes, these shares add up to more than 100% and should not be read as mutually exclusive categories.
Filing density is comparatively thin in areas like cross-project label-noise correction, patch-level classification, and control-flow-augmented AST embeddings relative to the volume of filings on core classifier and syntax-tree claims. These are technically specific gaps rather than a claim that the whole field is open, and a new filing there should still be checked against the most-cited existing records for overlapping claim language. Co-assignee data also shows very little cross-organization collaboration, just four pairs across the whole dataset, suggesting most technical development is happening inside single organizations rather than through joint filings.
The United States leads with 1,045 records, followed by WIPO PCT filings at 287 and the European Patent Office at 258. China follows at 232, with India at 124 and Canada at 59. This distribution reflects both where the leading assignees are headquartered and where they choose to pursue broader international protection through the PCT route.
Answers are grounded in the same dataset. Derived from a Patsnap search on Machine Learning for Vulnerability Detection 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 Machine Learning for Vulnerability Detection in depth with Eureka
Go past this page: query the whole machine learning for vulnerability detection corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
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.