Home › Guides › Anti-Bot vs Proxy Patents
✓ data: USPTO PatentsView grant and pre-grant publication tables (granted patents and published applications through 31 December 2025), plus USPTO Open Data Portal searches run on 2 October 2026 · method · caveats · download the data
Who is patenting the tools that stop bots, and who is patenting the proxy networks bots use to get around them? We searched the titles and abstracts of every US patent granted from 1976 to 2025 and every US patent application published from 2001 to 2025 for four topics: bot detection, browser and device fingerprinting, CAPTCHA, and residential proxies / IP rotation. Then we counted patents assigned to proxy-network companies. Every number on this page is recomputed by a script from saved copies of the USPTO data.
Read this first. These are keyword counts, so they are a sample, not a census. Hand checks put precision at about 90–96% per topic, but the search misses patents that describe the same ideas in other words. A granted patent is a right the USPTO issued. It isn’t proof the technology is in use, and the USPTO or a court can later narrow or cancel its claims. This page is not legal advice.
anti-bot keyword grants, all years 675in 2025 62Bright Data grants, 2012–2025 219residential-proxy / IP-rotation wording 9 grants
Headline numbers
- Anti-bot patenting is at a record. 62 granted US patents in 2025 matched our bot-detection, fingerprinting or CAPTCHA keywords, up from 25 in 2012. Across all years the total is 675.
- CAPTCHA patents peaked a decade ago. Fingerprinting took over. CAPTCHA grants peaked at 25 in 2014 and have run at 7–11 a year since 2020. Published CAPTCHA applications fell from 29 in 2017 to 5–10 a year in 2022–2025. Fingerprinting grants went from 3 in 2014 to 26 in 2025. Bot-detection grants peaked at 32 in 2022 and reached 27 in 2025.
- One proxy company out-patented the whole anti-bot keyword set in 2024. Bright Data (including its former name, Luminati Networks, and Hola Networks, where several of its patent families began) was granted 66 US patents in 2024. That year our keyword search found 52 anti-bot grants across all assignees. The comparison is lopsided by design, since a company count catches every patent and a keyword search misses some, but the scale is clear. In total Bright Data was granted 219 patents from 2012 to 2025.
- Most of those are continuations. Bright Data’s 219 grants carry only 22 distinct titles. 79 of them share the title “System and method for improving internet communication by using intermediate nodes”. Continuation patents reuse one original filing to claim more variations, so this measures portfolio strategy, not 219 separate inventions.
- The Oxylabs group holds 72 grants (2020–2025) under 29 distinct titles. The assignees are Oxylabs, UAB, Metacluster LT, UAB and Teso LT, UAB. It also has 62 published applications from 2021 to 2025, so more are in the pipeline.
- Almost nobody patents “residential proxies” by that name. Only 9 granted patents use residential-proxy or IP-rotation wording in their title or abstract. 3 are Fastly CDN patents and 3 are Oxylabs-group “smart proxy rotator” patents. None of Bright Data’s 219 grants uses those words (0 matches). They describe “intermediate nodes”, “tunnel devices” and “faster data communication” instead.
- Big Tech owns the anti-bot keyword patents. Akamai leads bot detection. The top owners across the three anti-bot topics are Microsoft (44), IBM (43), Amazon (27) and Akamai (26). Akamai has the most bot-detection grants (26).
- 2026 so far: bot detection is rising in new applications. A title-only search of the USPTO Open Data Portal finds 20 published applications from 1 January to 30 September 2026 with bot-detection titles. The same window found 17 in 2025 and 3 in 2016. These are small numbers from a narrower, title-only search, so they aren’t comparable with the totals above.
Anti-bot patents granted per year
Anti-bot keyword grants vs proxy-company grants
The numbers by year
Granted patents, by grant year
| Series | 2016 | 2017 | 2018 | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|---|---|---|---|---|
| Bot detection | 13 | 17 | 20 | 18 | 23 | 13 | 32 | 15 | 24 | 27 |
| Browser & device fingerprinting | 5 | 12 | 12 | 14 | 27 | 20 | 21 | 25 | 22 | 26 |
| CAPTCHA | 21 | 18 | 23 | 24 | 9 | 10 | 9 | 11 | 7 | 11 |
| Residential proxy / IP rotation | 0 | 0 | 1 | 0 | 4 | 0 | 0 | 1 | 1 | 2 |
| Any anti-bot topic | 35 | 46 | 52 | 54 | 58 | 41 | 61 | 51 | 52 | 62 |
| Assigned to Bright Data (incl. Luminati, Hola) | 3 | 2 | 0 | 5 | 11 | 30 | 23 | 34 | 66 | 38 |
| Assigned to Oxylabs group | 0 | 0 | 0 | 0 | 4 | 8 | 22 | 18 | 11 | 9 |
Published applications, by publication year. Applications are usually published about 18 months after filing, so this is the closest public signal of what was filed.
| Topic | 2016 | 2017 | 2018 | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|---|---|---|---|---|
| Bot detection | 22 | 9 | 17 | 18 | 25 | 13 | 25 | 14 | 17 | 23 |
| Browser & device fingerprinting | 9 | 10 | 18 | 15 | 24 | 25 | 24 | 21 | 21 | 25 |
| CAPTCHA | 26 | 29 | 11 | 11 | 14 | 9 | 7 | 5 | 6 | 10 |
| Residential proxy / IP rotation | 0 | 1 | 0 | 1 | 1 | 0 | 1 | 1 | 1 | 1 |
| Any anti-bot topic | 54 | 46 | 46 | 42 | 62 | 47 | 55 | 40 | 43 | 58 |
Granted patents by filing year are in the CSV. Filing years after 2021 are badly incomplete, because most of those applications haven’t been decided yet.
USPTO Open Data Portal, title-only search, published applications, 1 January–30 September of each year
| Title keywords | 2016 | 2017 | 2018 | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Bot detection | 3 | 5 | 7 | 10 | 12 | 5 | 7 | 4 | 9 | 17 | 20 |
| Fingerprinting | 3 | 2 | 4 | 8 | 10 | 11 | 7 | 8 | 5 | 7 | 4 |
| CAPTCHA | 12 | 18 | 6 | 4 | 5 | 3 | 4 | 3 | 4 | 4 | 3 |
| Residential proxy / IP rotation | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 |
Top patent owners (anti-bot topics, granted patents)
| # | Assignee (grouped) | Grants matching any anti-bot topic |
|---|---|---|
| 1 | Microsoft | 44 |
| 2 | IBM | 43 |
| 3 | Amazon | 27 |
| 4 | Akamai | 26 |
| 5 | Capital One | 22 |
| 6 | Shape Security (F5) | 13 |
| 7 | BioCatch | 12 |
| 8 | Cisco | 12 |
| 9 | 11 | |
| 10 | USAA | 11 |
By topic. Bot detection: Akamai (26), Microsoft (15), Amazon (14). Fingerprinting: Capital One (11), USAA (10), Cisco (10). CAPTCHA: Microsoft (34), IBM (25), Amazon (14). Residential proxy / IP rotation: Fastly (3), Oxylabs group (3), NxGen Partners IP, LLC (1).
Grouped: Microsoft Corporation + Microsoft Technology Licensing; Oxylabs, UAB + Metacluster LT, UAB + Teso LT, UAB. Shape Security is shown separately from its later parent F5. Full, ungrouped names are in the CSV.
Notable patents
Chosen to show the range of each topic, not ranked by importance. The descriptions are based on the patents’ titles and abstracts.
| Patent | Granted | Title | Assignee | Why it’s here |
|---|---|---|---|---|
| US 7,197,646 | 2007 | System and method for preventing automated programs in a network | Disney Enterprises, Inc. | Earliest grant in the bot-detection set (preventing automated programs from using a network service). |
| US 7,200,576 | 2007 | Secure online transactions using a captcha image as a watermark | Microsoft Corporation | Earliest grant in the CAPTCHA set. |
| US 8,280,993 | 2012 | System and method for detecting Internet bots | YAHOO HOLDINGS, INC. | An early patent with “detecting Internet bots” in its title. |
| US 8,601,109 | 2013 | Incremental browser-based device fingerprinting | BLUECAVA, INC. | Browser-based device fingerprinting, granted 2013. |
| US 9,986,058 | 2018 | Security systems for mitigating attacks from a headless browser executing on a client computer | Shape Security, Inc. | Detecting and mitigating headless browsers, the automation tool many scrapers use. |
| US 10,356,101 | 2019 | Utilizing transport layer security (TLS) fingerprints to determine agents and operating systems | PAYPAL, INC. | Using TLS fingerprints to tell what software a client is really running. |
| US 11,184,390 | 2021 | Bot detection in an edge network using transport layer security (TLS) fingerprint | Akamai Technologies, Inc. | Bot detection at a CDN edge using TLS fingerprints. |
| US 12,316,650 | 2025 | Computing system defenses to rotating IP addresses during computing attacks | PAYPAL, INC. | Defending against attackers who rotate IP addresses. |
| US 8,560,604 | 2013 | System and method for providing faster and more efficient data communication | Hola Networks Ltd. | Early grant from the Bright Data patent families (originally assigned to Hola Networks). |
| US 10,601,948 | 2020 | Smart proxy rotator | METACLUSTER LT, UAB | “Smart proxy rotator”: one of only 9 grants in our residential-proxy / IP-rotation keyword set. |
| US 11,722,577 | 2023 | Proxying TCP fingerprints | Webshare Software Company | From the proxy side: proxying TCP fingerprints. |
| US 11,818,104 | 2023 | Anonymous proxying | NetNut Ltd. | NetNut’s first grant in the data. |
Method
- Data. USPTO PatentsView bulk tables, downloaded on 2 October 2026 from the USPTO Open Data Portal (products PVGPATDIS and PVPGPUBDIS; PatentsView moved into the ODP in March 2026). We used patent and application titles and abstracts, grant, publication and filing dates, and PatentsView’s disambiguated assignee names. Coverage: grants and published applications through 31 December 2025. Only utility patents are counted; design and plant patents and withdrawn grants are excluded.
- Topics. We matched case-insensitive phrase rules against title and abstract text.
- Bot detection: detecting, blocking, distinguishing or mitigating bots, web robots, automated traffic or agents, headless browsers or non-human traffic; “bot detection”, “bot management”, “human or bot”. We excluded chatbots, conversational agents and robotic process automation, and documents about warehouse or mobile robots.
- Fingerprinting: browser, device, client, canvas, TLS/JA3, HTTP, TCP or user-agent fingerprints. We excluded biometric “fingerprint sensor” wording.
- CAPTCHA: CAPTCHA, reCAPTCHA, “completely automated public Turing test”, human interactive proofs, reverse Turing tests.
- Residential proxy / IP rotation: residential proxy, residential IP, IP, proxy or exit-node rotation, rotating proxies or IPs, proxy rotators.
- Hand checks. We read random samples of matches at each revision of the rules and tightened them until precision reached about 96% (bot detection, 24 of 25), 95% (CAPTCHA, 19 of 20) and 90% (fingerprinting, 18 of 20). Recall isn’t measured.
- Proxy companies. We counted every grant or application whose PatentsView disambiguated assignee is one of these names, whatever its wording:
- Bright Data: Bright Data Ltd.; Luminati Networks Ltd., its name until March 2021; Hola Networks Ltd.
- Oxylabs group: Oxylabs, UAB; Metacluster LT, UAB; Teso LT, UAB. Court filings describe Metacluster and Teso as Oxylabs-related entities, and they share patent titles with Oxylabs, UAB.
- NetNut: NetNut Ltd.
- Webshare: Webshare Software Company.
- 2026. PatentsView stops at 2025, so for 2026 we used the USPTO Open Data Portal search API. It runs title-only queries, counted for 1 January–30 September of each year so the years are comparable. Every request and response is saved.
- Verification. A script reruns the topic rules on the saved candidate texts and re-reads the assignee table and the saved API responses. It checks that every number on this page matches. Links go to Google Patents, which also shows each patent’s family, legal status and later reassignments.
Caveats
- Keyword counts undercount. Many anti-bot patents talk about “fraud”, “risk scoring” or “anomalous clients” and never use our phrases. Treat the topic totals as a lower bound and look at the trends, not the absolute levels.
- Company counts and keyword counts measure different things. The proxy-company series includes every patent those assignees hold, so the 2024 comparison overstates the gap.
- Recent filing years are incomplete. Applications are secret for about 18 months, and grants take years. Use grant or publication year for trends.
- Assignee names are as recorded. They reflect the assignee on the patent as printed, as cleaned up by PatentsView. Later sales or name changes may not show. Applications often list no assignee at publication, so the published-application company counts (Bright Data: 5) are far below the grant counts.
- Counts aren’t quality. A count of granted patents says nothing about whether any claim is valid, enforced or used. Several patents in these families have been challenged at the USPTO or in court. We do not assess any of them.
- US only. We count US patents only, not EPO, China or PCT filings.
Download the data
- Every matched patent and application (CSV, 979 granted patents + 841 published applications). Columns: id, type, grant or publication date, filing date, title, topic flags, proxy-company flag, assignees, the phrase that matched, and a Google Patents link.
- Counts by year (CSV): every series on this page, plus the ODP 2026 queries.
Cite as: ProxyPicker, “Anti-Bot vs Proxy Patents”, 2 October 2026, https://proxypicker.com/anti-bot-vs-proxy-patents/. Underlying data: USPTO PatentsView and the USPTO Open Data Portal.
Related: Anti-Bot Census (how many websites screen visitors for bots) · Residential Proxy Botnet & Takedown Timeline · What’s my IP? Proxy & VPN checker · Best proxies for web scraping · The Scraping Workforce (who does scraping work and what they earn) · reviews: Bright Data · Oxylabs · Webshare · NetNut.
Press & data requests: email [email protected] for interviews, the raw data behind this page, or corrections. See also our contact page.