Daily · data as of 16 Nov 2026
The state of global vulnerability research, from every public source, joined.
407 052 published CVEs. 35 655 with a public exploit. 1 802 known to be exploited in the wild. One record per CVE, followed from the day it was reserved to the day it was patched or attacked.
Key findings
What the joined data shows
- 01 Volume CVEs published per year
98 175 CVEs so far in 2026, +137% on the same period of 2025. At this pace the year ends near 111 981.
- 02 Exploitation Days from publication to first 'exploited' listing
Half of the 914 exploited CVEs published since November 2021 were listed within 7 days. 30% were already being exploited when published (zero-days), 51% within a week.
- 03 Exploitation Did a public exploit come first?
Of 916 exploited CVEs (published since November 2021), 39% never had a public exploit: attackers did not need one.
- 04 Prediction Effort against coverage: EPSS or CVSS
From the scores of November 2025: patching every CVE with EPSS ≥ 10% meant 6.9% of CVEs and caught 49% of those exploited in the next year. Patching every CVSS ≥ 9 meant 13% of CVEs and caught 31%. A small sample: 75 CVEs scored that day were first listed in the following year.
- 05 Records What NVD did with each month's CVEs
18 172 CVEs are waiting for NVD analysis and 63 265 are deferred. Of the last 12 months' CVEs, NVD has analysed 44%.
- 06 Records CNA score minus NVD score
Where both scored (since 2020, 80 816 CVEs), they agree exactly 25% of the time. NVD is at least 2 points higher in 21% of cases; the CNA in 6%.
- 07 Who Most credited researchers, all time
25 620 distinct names are credited in CVE records. 58% of them appear once; the top 1% of names account for 38% of credits.
- 08 Open source Malicious packages reported per month
241 648 malicious packages are recorded in OSV, 93% of them on npm. These are not bugs: they are packages published to steal or attack.
- 09 Exploitation Most claimed proofs of concept on GitHub
CVEs with ten or more PoC repositories are exploited in 51% of cases; with one to nine, 6%. Many repositories are fakes or scanners.
- 10 Weaknesses Most common weaknesses, last 12 months
The most common is CWE-79: Improper Neutralization of Input During Web Page Generation. 6% of the year's CVEs name no CWE.
This month
What is new
Updated every day from the sources below. Every CVE links to its full record.
# Newly listed as exploited
34 CVEs were first listed as exploited in the last 30 days.
| CVE-2026-114713 | 13 Nov 2026 | 5 | 8.6 | 0.51% | The labeled-response module in UnrealIRCd 6.0.0 before 6.2.7.1 does not escape the IRCv3 "label" message tag w |
| CVE-2026-123039 | 13 Nov 2026 | 1 | 7.8 | 2.2% | Local privilege escalation to root in the DynamicCLI daemon via local world-writable control socket |
| CVE-2026-111560 | 10 Nov 2026 | 8 | 8.6 | 9.1% | OS Command Injection in Barracuda Appliances |
| CVE-2026-111793 | 10 Nov 2026 | 25 | 4.3 | 0.64% | Openwave Webtop Media Transcoder does not enforce administrator-defined file type restrictions to restrict the |
| CVE-2026-116606 | 10 Nov 2026 | 8 | 9.3 | 11% | Authentication Bypass Allows Global Configuration Restore in Barracuda Email Security Gateway |
| CVE-2026-124194 | 10 Nov 2026 | 0 | 6.5 | 1.6% | Microsoft Outlook Spoofing Vulnerability |
| CVE-2026-125350 | 10 Nov 2026 | 0 | 8.8 | 0.96% | Use after free in XML in Google Chrome prior to 157.0.8103.50 allowed a remote attacker to execute arbitrary c |
| CVE-2026-20320 | 10 Nov 2026 | 83 | 7.5 | 0.98% | A vulnerability in the Open Client Interface (OCI) XML Parser of Cisco BroadWorks could allow an unauthenticat |
| CVE-2026-102895 | 9 Nov 2026 | 5 | 9.8 | 0.52% | GPUStack: Authentication bypass via trusted loopback source (reverse-proxy TCP peer or X-GPUStack-Real-IP header) grants unauthenticated admin |
| CVE-2026-116779 | 9 Nov 2026 | 4 | 9.8 | 2.5% | FreeSWITCH: Inbound SIP MESSAGE can select the api chat proto and run FreeSWITCH API commands |
| CVE-2026-123158 | 9 Nov 2026 | 4 | 8.1 | 19% | sshd in OpenSSH before 10.6p2 on Linux, when musl libc is used without PAM, lacks a workaround for a password |
| CVE-2026-123218 | 9 Nov 2026 | 2 | 9.8 | 0.53% | IBM Aspera Faspex Authorization Bypass |
# Most new PoC repositories, last 30 days
What people are building exploits for, by GitHub repositories created in the last 30 days.
| CVE-2026-109618 | 778 | Yes | 100% | RCE via OOM corruption |
| CVE-2026-111497 | 93 | Yes | 83% | Out-of-Bounds Write in Google Brotli Decoder |
| CVE-2026-98432 | 82 | no | 0.78% | ipv4: free inet_opt and ireq_opt after an RCU grace period |
| CVE-2026-98260 | 62 | no | 0.19% | exec: Cleanup POSIX timers right after de_thread() |
| CVE-2021-44228 | 52 | Yes | 100% | Apache Log4j2 JNDI features do not protect against attacker controlled LDAP and other JNDI related endpoints |
| CVE-2026-123158 | 40 | Yes | 19% | sshd in OpenSSH before 10.6p2 on Linux, when musl libc is used without PAM, lacks a workaround for a password |
| CVE-2026-43499 | 36 | no | 5.4% | rtmutex: Use waiter::task instead of current in remove_waiter() |
| CVE-2026-109619 | 30 | Yes | 100% | RCE via stale global state |
| CVE-2026-119618 | 16 | no | 1.6% | Apache HTTP Server: mod_http2: heap corruption and RCE in assign_headers |
| CVE-2026-99999 | 13 | no | 0.79% | ipv4: validate checksum_start before completing checksum |
# Biggest EPSS rises in 30 days
CVEs whose predicted chance of exploitation rose most in a month.
| CVE-2026-75650 | 4.0% | 99% | Yes | Adobe Commerce | Improper Neutralization of Special Elements Used in a Template Engine (CWE-1336) |
| CVE-2026-31431 | 3.4% | 98% | Yes | crypto: algif_aead - Revert to operating out-of-place |
| CVE-2026-34486 | 6.6% | 99% | Yes | Apache Tomcat: Fix for CVE-2026-29146 allowed bypass of EncryptInterceptor |
| CVE-2026-43500 | 2.3% | 94% | no | rxrpc: Also unshare DATA/RESPONSE packets when paged frags are present |
| CVE-2026-3055 | 4.0% | 91% | Yes | Insufficient input validation leading to memory overread |
| CVE-2026-28318 | 1.9% | 87% | Yes | SolarWinds Serv-U Unauthenticated Denial of Service Vulnerability |
| CVE-2026-35616 | 9.1% | 94% | Yes | A improper access control vulnerability in Fortinet FortiClientEMS 7.4.5 through 7.4.6 may allow an unauthenti |
| CVE-2026-42208 | 5.8% | 89% | Yes | LiteLLM: SQL injection in Proxy API key verification |
| CVE-2026-105133 | 0.85% | 84% | Yes | Ahsay AhsayCBS API ApiStructsAction.java checkSysPwd improper authentication |
| CVE-2026-34197 | 15% | 98% | Yes | Apache ActiveMQ Broker, Apache ActiveMQ All, Apache ActiveMQ: Authenticated users could perform RCE via Jolokia MBeans |
# Newest critical CVEs
1 439 CVEs scored critical in the last 30 days.
| CVE-2026-93664 | 16 Nov 2026 | 10.0 | 3DS | Deserialization of Untrusted Data vulnerability affecting Station Launcher App in 3DEXPERIENCE platform from Release 3DEXPERIENCE R2023x through Release 3DEXPERIENCE R2026x |
| CVE-2026-53998 | 16 Nov 2026 | 10.0 | GitHub_M | Phoenix: Format String Injection in applyChatTemplate |
| CVE-2026-129354 | 15 Nov 2026 | 10.0 | VulDB | D-Link DIR-600L HNAP cgibin hnap_main stack-based overflow |
| CVE-2026-129444 | 15 Nov 2026 | 10.0 | VulDB | D-Link DSL-6740C FTP Daemon ftpd_conf.php hard-coded credentials |
| CVE-2026-129353 | 14 Nov 2026 | 10.0 | VulDB | D-Link DIR-600L authentication.cgi authenticationcgi_main stack-based overflow |
| CVE-2026-129351 | 14 Nov 2026 | 10.0 | VulDB | TRENDnet TEW-657BRM mini_httpd setup.cgi sprintf out-of-bounds write |
| CVE-2026-77033 | 16 Nov 2026 | 9.9 | 3DS | OS Command Injection vulnerability affecting Station Launcher App in 3DEXPERIENCE platform from Release 3DEXPERIENCE R2023x through Release 3DEXPERIENCE R2026x |
| CVE-2026-19732 | 16 Nov 2026 | 9.9 | 3DS | OS Command Injection vulnerability affecting Station Launcher App in 3DEXPERIENCE platform from Release 3DEXPERIENCE R2023x through Release 3DEXPERIENCE R2026x |
| CVE-2026-127048 | 15 Nov 2026 | 9.8 | WPScan | Lazy Blocks < 3.0.0 - Unauthenticated RCE |
| CVE-2026-125795 | 15 Nov 2026 | 9.8 | Patchstack | WordPress Estatik plugin <= 4.3.6 - Local File Inclusion vulnerability |
The whole census
Everything, on this one page
All eight reports follow, one after the other. Each also has its own page in the menu above.
566 CNAs · 25 566 vendors · 25 620 credited researchers, each with a page.
How many vulnerabilities get a CVE, and how fast that is growing
Every published CVE record since 1999, by publication date. Rejected IDs are counted separately; reserved IDs that were never published are not visible.
- Published CVEs, all time
- 407 052
- So far in 2026
- 98 175
- Same period last year
- 41 459+137%
- 2026 at this pace
- 111 981
Severity
# Severity mix by year
Share of each year's CVEs by severity: the CNA's score where given, else CISA's, else NVD's.
- Critical
- High
- Medium
- Low
- No score
Process
From an ID being reserved to the record being published; withdrawn IDs; how many organisations take part, and how concentrated publishing is.
# Days from reserving an ID to publishing it
Reserving comes when a CNA starts handling a report; publishing after the fix. The gap is the embargo, roughly.
- Median
- 90th percentile
# Published and rejected IDs, by ID year
Rejected IDs were reserved and later withdrawn: duplicates, not vulnerabilities, or never used.
- Published
- Rejected
# How many CNAs publish
Distinct CNAs that published at least one CVE that year.
# How concentrated publishing is
Share of each year's CVEs published by the largest CNA, and by the ten largest.
- Ten largest CNAs
- Largest CNA
From publication to public exploit to attack
When CVEs get a public exploit, when they are seen exploited in the wild, and which come first. 'Exploited' means listed by CISA KEV, ENISA's EUVD, VulnCheck KEV or Google Project Zero; nothing else.
- Exploited CVEs since Nov 2021
- 914
- Median days to first listing
- 7
- Exploited by publication day
- 30%
- Exploited without any public exploit
- 39%
The funnel
# From published to exploited
Of 243 415 CVEs published since November 2021, 3.8% have a public exploit or proof of concept and 0.40% are known to be exploited in the wild.
Bar lengths on a logarithmic scale.
Timing
CVEs published since 3 November 2021, when CISA’s list started; older CVEs were added to it in bulk, which would distort the timing.
# Days from publication to first 'exploited' listing
Half of the 914 exploited CVEs published since November 2021 were listed within 7 days. 30% were already being exploited when published (zero-days), 51% within a week.
- Before publication
- Same day
- After
# Days from publication to first public exploit
For 8 729 CVEs published since November 2021 with a public exploit, the median wait was 5 days; 27% had one before the CVE was published.
- Before publication
- Same day
- After
# Did a public exploit come first?
Of 916 exploited CVEs (published since November 2021), 39% never had a public exploit: attackers did not need one.
# Median days to first exploited listing, by publication year
Recent years look faster partly because slow cases have not happened yet.
Zero-days
# Exploited CVEs per year, zero-day or not
By year of first listing; CVEs published since November 2021.
- Exploited on or before publication
- Exploited after publication
# Project Zero: zero-days caught in the wild
Google Project Zero's tracker. Where it records both dates, the median time from discovery to patch was 9 days (132 cases).
# Whose software the in-the-wild zero-days hit
Vendors in Project Zero's in-the-wild list, all years.
# CISA KEV additions per year, by ransomware use
CISA marks KEV entries known to be used in ransomware campaigns.
- Known ransomware use
- No known ransomware use
The lists
# Who says it is exploited
CISA lists 1 781 CVEs, ENISA 1 798. ENISA's list includes 1 781 of CISA's and adds 17 of its own; 0 are on CISA's list only.
Proof-of-concept hype
# Most claimed proofs of concept on GitHub
CVEs with ten or more PoC repositories are exploited in 51% of cases; with one to nine, 6%. Many repositories are fakes or scanners.
| CVE-2026-109618 | 778 | Yes | 100% | RCE via OOM corruption |
| CVE-2021-44228 | 509 | Yes | 100% | Apache Log4j2 JNDI features do not protect against attacker controlled LDAP and other JNDI related endpoints |
| CVE-2025-55182 | 459 | Yes | 100% | – |
| CVE-2026-31431 | 324 | Yes | 98% | crypto: algif_aead - Revert to operating out-of-place |
| CVE-2026-43499 | 194 | no | 5.4% | rtmutex: Use waiter::task instead of current in remove_waiter() |
| CVE-2021-4034 | 184 | Yes | 95% | – |
| CVE-2021-41773 | 176 | Yes | 100% | Path traversal and file disclosure vulnerability in Apache HTTP Server 2.4.49 |
| CVE-2025-29927 | 134 | no | 99% | Authorization Bypass in Next.js Middleware |
| CVE-2019-0708 | 122 | Yes | 100% | – |
| CVE-2022-0847 | 110 | Yes | 93% | – |
| CVE-2022-22965 | 100 | Yes | 100% | – |
| CVE-2024-6387 | 97 | no | 100% | Openssh: regresshion - race condition in ssh allows rce/dos |
| CVE-2014-6271 | 96 | Yes | 100% | – |
| CVE-2022-30190 | 95 | Yes | 98% | Microsoft Windows Support Diagnostic Tool (MSDT) Remote Code Execution Vulnerability |
| CVE-2026-111497 | 93 | Yes | 83% | Out-of-Bounds Write in Google Brotli Decoder |
Did the scores see it coming?
EPSS gives every CVE a daily probability of exploitation in the next 30 days; CVSS rates how bad a CVE would be. Both are used to decide what to patch first. Here, past scores are checked against what was later listed as exploited.
- CVEs scored by EPSS today
- 406 741
- With EPSS ≥ 10%
- 17 550
- With EPSS ≥ 50%
- 4 384
Effort against coverage
If you had patched every CVE above a threshold, how much work would that have been, and how many of the CVEs exploited in the following year would you have caught? The dashed line is choosing at random.
# Effort against coverage: EPSS or CVSS
From the scores of November 2025: patching every CVE with EPSS ≥ 10% meant 6.9% of CVEs and caught 49% of those exploited in the next year. Patching every CVSS ≥ 9 meant 13% of CVEs and caught 31%. A small sample: 75 CVEs scored that day were first listed in the following year.
- EPSS at or above a threshold
- CVSS at or above a score
- Random choice
Calibration
# Predicted against observed, 90 days ahead
Each first-of-month snapshot since 2022, grouped by score. 'Observed' is the share first listed as exploited in the next 90 days. Of scores of 60% or more, 0.20% were followed by a first listing within 90 days, against 83% predicted. EPSS predicts exploitation activity of any kind, such as attempts seen by sensors; the lists record confirmed exploitation, mostly of new CVEs, and CVEs already listed are left out here. Read it as a ranking, not a probability of being listed.
- Predicted (mean EPSS)
- Observed (listed as exploited within 90 days)
| EPSS | Scores | Predicted | Listed within 90 days | Observed |
|---|---|---|---|---|
| 0–0.001 | 3 497 258 | 0.06% | 235 | 0.01% |
| 0.001–0.01 | 5 742 795 | 0.43% | 324 | 0.01% |
| 0.01–0.05 | 2 662 064 | 1.95% | 532 | 0.02% |
| 0.05–0.1 | 447 310 | 7.10% | 134 | 0.03% |
| 0.1–0.3 | 390 473 | 17.41% | 208 | 0.05% |
| 0.3–0.6 | 167 622 | 42.96% | 224 | 0.13% |
| 0.6–1 | 217 694 | 82.62% | 439 | 0.20% |
Before the listing
# EPSS just before a CVE was listed as exploited
The last available score before listing, for 1 145 CVEs listed more than a week after publication: 43% were already at 10% or more; the median was 5.1%.
Who assigns, who ships the bugs, who finds them
CVE Numbering Authorities assign the IDs; vendors are named in the records; researchers are credited by the CNA. Every name links to its own page with all its CVEs.
- CNAs
- 566
- Published in the last 12 months
- 426
- Vendor names
- 25 566
- Credited names
- 25 620
CVE Numbering Authorities
# CNAs: volume and record quality
566 organisations may assign CVE IDs; 426 published at least one in the last 12 months. The busiest was GitHub, Inc. with 13 590.
| GitHub, Inc. | USA | 13 590 | 22 755 | 49 | |||
| VulnCheck | USA | 11 172 | 11 830 | 21 | |||
| kernel.org | USA | 9 343 | 18 245 | 12 | |||
| VulDB | Switzerland | 7 034 | 18 029 | 6 | |||
| Patchstack | Estonia | 6 430 | 19 644 | 0 | |||
| Microsoft Corporation | USA | 5 608 | 17 053 | 384 | |||
| Wordfence | USA | 4 553 | 12 157 | 1 | |||
| MITRE Corporation | USA | 4 539 | 115 582 | 332 | |||
| Oracle | USA | 4 154 | 12 397 | 44 | |||
| Chrome | USA | 3 998 | 7 447 | 78 | |||
| WPScan | France | 2 367 | 6 524 | 2 | |||
| IBM Corporation | USA | 1 909 | 8 441 | 10 | |||
| Apache Software Foundation | USA | 1 517 | 3 245 | 35 | |||
| Red Hat, Inc. | USA | 1 448 | 12 112 | 37 | |||
| Apple Inc. | USA | 1 270 | 7 732 | 96 | |||
| Adobe Systems Incorporated | USA | 1 269 | 7 898 | 77 | |||
| Mozilla Corporation | USA | 984 | 3 159 | 12 | |||
| Dell | USA | 701 | 2 618 | 3 | |||
| Cisco Systems, Inc. | USA | 604 | 6 653 | 102 | |||
| Cybersecurity and Infrastructure Security Agency (CISA) Industrial Control Systems (ICS) | USA | 569 | 3 911 | 8 | |||
| GitLab Inc. | USA | 498 | 1 513 | 7 | |||
| Hewlett Packard Enterprise (HPE) | USA | 474 | 1 715 | 2 | |||
| Android (associated with Google Inc. or Open Handset Alliance) | USA | 457 | 5 396 | 14 | |||
| CPAN Security Group | Canada | 439 | 484 | 0 | |||
| NVIDIA Corporation | USA | 432 | 1 209 | 0 |
Countries and kinds of CNA
# Where CVEs come from: CNA country
CVEs published in the last 12 months, by the country of the CNA that assigned them (not of the vendor or researcher).
# CNAs per country
Organisations authorised to assign CVE IDs, by country.
# What kind of organisations CNAs are
CNA types as listed by the CVE Program (one CNA can have several).
Researchers
# Most credited researchers, all time
25 620 distinct names are credited in CVE records. 58% of them appear once; the top 1% of names account for 38% of credits.
| WPScan | 4 739 | 2 364 | 277 | 0 | 5 Apr 2021 | 16 Nov 2026 |
| VulDB CNA Team | 2 747 | 2 715 | 161 | 0 | 28 Apr 2023 | 15 Nov 2026 |
| João Pedro S Alcântara (Kinorth) | Patchstack Bug Bounty Program | 1 056 | 436 | 73 | 0 | 9 Jul 2024 | 23 Oct 2026 |
| SOPROBRO | Patchstack Bug Bounty Program | 711 | 1 | 3 | 0 | 5 Oct 2024 | 5 Jan 2026 |
| Tran Nguyen Bao Khanh (VCI - VNPT Cyber Immunity) | Patchstack Bug Bounty Program | 697 | 496 | 89 | 0 | 24 Jan 2025 | 15 Nov 2026 |
| George Chen | 626 | 626 | 76 | 0 | 19 May 2026 | 16 Nov 2026 |
| Nabil Irawan | Patchstack Bug Bounty Program | 624 | 274 | 6 | 0 | 24 Feb 2025 | 19 Aug 2026 |
| Francesco Carlucci | 565 | 1 | 16 | 0 | 8 Nov 2021 | 9 Nov 2026 |
| VulDB GitHub Commit Analyzer | 510 | 15 | 0 | 0 | 2 Nov 2022 | 9 Nov 2026 |
| Rafie Muhammad (Patchstack) | 502 | 0 | 91 | 0 | 22 Nov 2022 | 28 Aug 2025 |
| Bonds | Patchstack Bug Bounty Program | 476 | 322 | 101 | 0 | 17 Sep 2024 | 30 Oct 2026 |
| István Márton | 471 | 3 | 76 | 0 | 14 Feb 2023 | 27 Nov 2025 |
| Nguyen Xuan Chien | Patchstack Bug Bounty Program | 435 | 31 | 4 | 0 | 28 Feb 2024 | 2 Jun 2026 |
| Dmitrii Ignatyev | 419 | 108 | 0 | 0 | 30 Aug 2023 | 5 Nov 2026 |
| Krzysztof Zając | 418 | 2 | 30 | 0 | 29 Nov 2021 | 16 Aug 2026 |
| Mika | Patchstack Bug Bounty Program | 403 | 13 | 27 | 0 | 10 Apr 2024 | 17 Jun 2026 |
| Mat Powell of Trend Micro Zero Day Initiative | 380 | 3 | 0 | 0 | 3 Jun 2019 | 19 Dec 2025 |
| Wordfence PRISM | 372 | 372 | 5 | 0 | 23 Jul 2026 | 13 Nov 2026 |
| VulnCheck | 371 | 371 | 49 | 0 | 18 Dec 2025 | 26 Oct 2026 |
| Peter Thaleikis | 362 | 18 | 4 | 0 | 7 Mar 2024 | 31 Oct 2026 |
| Muhammad Yudha - DJ | Patchstack Bug Bounty Program | 360 | 71 | 0 | 0 | 24 Jan 2025 | 8 Oct 2026 |
| Muhammad Yudha - DJ | 353 | 172 | 3 | 0 | 1 Oct 2024 | 10 Oct 2026 |
| hieuPenguinnn | 352 | 352 | 29 | 0 | 11 Oct 2026 | 16 Nov 2026 |
| Bob Matyas | 348 | 8 | 3 | 0 | 19 Jun 2023 | 2 Sep 2026 |
| Lucio Sá | 338 | 1 | 20 | 0 | 3 Jan 2024 | 10 Jul 2026 |
| Lana Codes | 335 | 0 | 1 | 0 | 17 Jul 2022 | 16 Jan 2024 |
| SSL_Seven_Security Lab_WangZhiQiang_XiaoZiLong (VulDB User) | 330 | 17 | 0 | 0 | 15 Jan 2023 | 20 Sep 2026 |
| Greg Knaddison (greggles) | 329 | 255 | 14 | 1 | 31 Mar 2025 | 30 Oct 2026 |
| Phat RiO | Patchstack Bug Bounty Program | 327 | 254 | 58 | 0 | 7 Feb 2025 | 22 Oct 2026 |
| Rafael Pedrero | 327 | 36 | 73 | 0 | 9 Mar 2022 | 21 Apr 2026 |
What kind of bugs: weaknesses and techniques
CWE names the kind of mistake behind a vulnerability; MITRE ATT&CK names what an attacker gains by exploiting it. Each CWE links to its page with every CVE that names it.
Rising and falling
# Rising and falling weaknesses
The year's top 25 against the 12 months before.
| 1 | CWE-79 | Improper Neutralization of Input During Web Page Generation | 9 559 | +13% | = |
| 2 | CWE-862 | Missing Authorization | 5 971 | +159% | ▲ 2 |
| 3 | CWE-284 | Improper Access Control | 4 813 | +296% | ▲ 5 |
| 4 | CWE-89 | Improper Neutralization of Special Elements used in an SQL Command | 4 678 | +20% | ▼ 2 |
| 5 | CWE-416 | Use After Free | 3 694 | +214% | ▲ 4 |
| 6 | CWE-22 | Improper Limitation of a Pathname to a Restricted Directory | 3 377 | +231% | ▲ 6 |
| 7 | CWE-125 | Out-of-bounds Read | 2 816 | +158% | ▲ 3 |
| 8 | CWE-200 | Exposure of Sensitive Information to an Unauthorized Actor | 2 667 | +235% | ▲ 10 |
| 9 | CWE-863 | Incorrect Authorization | 2 626 | +448% | ▲ 16 |
| 10 | CWE-639 | Authorization Bypass Through User-Controlled Key | 2 503 | +617% | ▲ 22 |
| 11 | CWE-78 | Improper Neutralization of Special Elements used in an OS Command | 2 417 | +188% | ▲ 3 |
| 12 | CWE-918 | Server-Side Request Forgery | 2 398 | +353% | ▲ 12 |
| 13 | CWE-122 | Heap-based Buffer Overflow | 2 306 | +316% | ▲ 10 |
| 14 | CWE-20 | Improper Input Validation | 2 271 | +250% | ▲ 6 |
| 15 | CWE-94 | Improper Control of Generation of Code | 2 261 | +63% | ▼ 9 |
| 16 | CWE-74 | Improper Neutralization of Special Elements in Output Used by a Downstream Component | 2 238 | −7% | ▼ 13 |
| 17 | CWE-787 | Out-of-bounds Write | 2 233 | +153% | ▼ 4 |
| 18 | CWE-306 | Missing Authentication for Critical Functionality | 2 103 | +413% | ▲ 11 |
| 19 | CWE-400 | Uncontrolled Resource Consumption | 1 643 | +270% | ▲ 7 |
| 20 | CWE-121 | Stack-based Buffer Overflow | 1 627 | +95% | ▼ 5 |
| 21 | CWE-119 | Improper Restriction of Operations within the Bounds of a Memory Buffer | 1 498 | +38% | ▼ 10 |
| 22 | CWE-502 | Deserialization of Untrusted Data | 1 495 | +140% | ▼ 1 |
| 23 | CWE-269 | Improper Privilege Management | 1 490 | +354% | ▲ 11 |
| 24 | CWE-287 | Improper Authentication | 1 487 | +314% | ▲ 7 |
| 25 | CWE-352 | Cross-Site Request Forgery | 1 485 | −23% | ▼ 20 |
Families over time
# Three families of weakness over time
Share of each year's CVEs (among those naming a CWE) in each family. A CVE can be in more than one.
- Injection (XSS, SQL, command, code)
- Memory safety (overflow, use-after-free, …)
- Access control and authentication
# The same families among exploited CVEs
By year of first 'exploited' listing.
- Injection
- Memory safety
- Access control
How good the records are, and what NVD does with them
A CVE record is only as useful as what is in it: a severity, a weakness, the affected versions. CNAs write the record; CISA and NIST's NVD add to it. This is how complete records are and how far behind the enrichment runs.
- Waiting for NVD analysis
- 18 172
- Deferred by NVD
- 63 265
- Analysed, last 12 months
- 44%
NVD
# What NVD did with each month's CVEs
18 172 CVEs are waiting for NVD analysis and 63 265 are deferred. Of the last 12 months' CVEs, NVD has analysed 44%.
- Analysed
- Modified after analysis
- Deferred (will not be analysed)
- Waiting
- Other or not in NVD
Completeness
# How complete CNA records are
Share of each year's CVE records in which the CNA itself supplied each field. Missing fields are what NVD and CISA fill in, when they can.
- Has a CVSS score
- Names a weakness (CWE)
- Lists affected versions
- Two or more references
# Who scored the CVE
The first available CVSS score, in this order: the CNA's, CISA's, NVD's.
- The CNA
- CISA (ADP)
- NVD only
- Nobody
CISA enrichment
# CISA's enrichment (Vulnrichment)
Share of each month's CVEs carrying CISA's ADP container, and an SSVC exploitation decision.
- CISA enrichment present
- CISA SSVC decision present
Package ecosystems, malicious packages and distribution fixes
OSV merges the advisories of every open-source ecosystem: GitHub's, PyPA's, Go's, RustSec's, and the Linux distributions'. It also records packages published to attack, which are not vulnerabilities at all.
- Malicious packages recorded
- 241 648
- Linux kernel share of CVEs, last year
- 8.9%
- Median days, CVE to OSV advisory
- 0
Ecosystems
# Open-source advisories per year, by ecosystem
Security advisories (GitHub, PyPA, Go, RustSec and others, merged by OSV) for packages in the main language ecosystems. Malicious packages excluded.
- npm
- PyPI
- Maven
- Go
- Packagist
- crates.io
- NuGet
- RubyGems
# Ecosystems at a glance
| PyPI | 14 276 | 7 036 | 13 948 | 12 034 | |
| Go | 9 755 | 4 794 | 9 069 | 20 | |
| npm | 7 740 | 3 589 | 6 669 | 224 100 | |
| Packagist | 7 300 | 1 894 | 6 440 | 1 | |
| Maven | 7 133 | 1 115 | 7 037 | 2 | |
| crates.io | 2 999 | 1 086 | 1 808 | 25 | |
| NuGet | 1 175 | 440 | 1 118 | 782 | |
| RubyGems | 1 142 | 246 | 1 097 | 4 321 | |
| Hex | 436 | 393 | 436 | 0 | |
| GitHub Actions | 63 | 25 | 48 | 0 | |
| Pub | 12 | 2 | 11 | 0 |
Malicious packages
# Malicious packages reported per month
241 648 malicious packages are recorded in OSV, 93% of them on npm. These are not bugs: they are packages published to steal or attack.
- npm
- PyPI
- Other ecosystems
Packages
# Packages with the most advisories, last 12 months
Linux distributions
# Days from CVE publication to the first fixed distro advisory
CVEs published since September 2024 that Debian, Ubuntu or Red Hat later shipped a fix for, through a security advisory in OSV.
| Debian | 28 311 | 0 | 50 | 0 |
| Ubuntu | 17 089 | 0 | 33 | 0 |
| Red Hat | 7 700 | 21 | 31 | 23 |
# Distribution of days to a distro fix
- Before publication
- Same day
- After
The kernel
# The Linux kernel, a CNA since 2024
The kernel team assigns a CVE to most fixes that could matter: 18 245 so far, 9% of all CVEs in the last 12 months.
Papers, conferences and bug bounties
Security research beyond CVE records: what is written on arXiv and presented at the top four academic security conferences, which CVEs papers talk about, and what bug-bounty programmes choose to disclose.
Papers
# Security papers on arXiv per year
7 912 papers in arXiv's Cryptography and Security category (cs.CR) in 2025, +21% on 2024.
# Papers at the 'big four' security conferences
1 299 papers at the four top academic security conferences in 2025; 1 237 in 2024.
- IEEE S&P
- USENIX Security
- ACM CCS
- NDSS
# Most prolific authors at the big four, last five years
# Papers naming CVEs, per year
CVEs in papers
# CVEs most often named in papers
40 cs.CR papers name at least one CVE in their abstract.
| CVE-2025-32711 | 2 | no | M365 Copilot Information Disclosure Vulnerability |
| CVE-2024-25825 | 1 | no | FydeOS for PC 17.1 R114, FydeOS for VMware 17.0 R114, FydeOS for You 17.1 R114, and OpenFyde R114 were discove |
| CVE-2024-41623 | 1 | no | An issue in D3D Security D3D IP Camera (D8801) v.V9.1.17.1.4-20180428 allows a local attacker to execute arbit |
| CVE-2026-6297 | 1 | no | Use after free in Proxy in Google Chrome prior to 147.0.7727.101 allowed an attacker in a privileged network p |
| CVE-2023-50868 | 1 | no | The Closest Encloser Proof aspect of the DNS protocol (in RFC 5155 when RFC 9276 guidance is skipped) allows r |
| CVE-2017-5638 | 1 | Yes | The Jakarta Multipart parser in Apache Struts 2 2.3.x before 2.3.32 and 2.5.x before 2.5.10.1 has incorrect ex |
| CVE-2019-19926 | 1 | no | multiSelect in select.c in SQLite 3.30.1 mishandles certain errors during parsing, as demonstrated by errors f |
| CVE-2024-42531 | 1 | no | Ezviz Internet PT Camera CS-CV246 D15655150 allows an unauthenticated host to access its live video stream by |
| CVE-2020-15802 | 1 | no | Devices supporting Bluetooth before 5.1 may allow man-in-the-middle attacks, aka BLURtooth. Cross Transport Ke |
| CVE-2023-38545 | 1 | no | This flaw makes curl overflow a heap based buffer in the SOCKS5 proxy handshake. When curl is asked to pass a |
| CVE-2025-6021 | 1 | no | Libxml2: integer overflow in xmlbuildqname() leads to stack buffer overflow in libxml2 |
| CVE-2026-25253 | 1 | no | OpenClaw (aka clawdbot or Moltbot) before 2026.1.29 obtains a gatewayUrl value from a query string and automat |
| CVE-2022-46480 | 1 | no | Incorrect Session Management and Credential Re-use in the Bluetooth LE stack of the Ultraloq UL3 2nd Gen Smart |
| CVE-2025-56538 | 1 | – | – |
| CVE-2026-5280 | 1 | no | Use after free in WebCodecs in Google Chrome prior to 146.0.7680.178 allowed a remote attacker to execute arbi |
| CVE-2021-33813 | 1 | no | An XXE issue in SAXBuilder in JDOM through 2.0.6 allows attackers to cause a denial of service via a crafted H |
| CVE-2023-20570 | 1 | no | Insufficient verification of data authenticity in the configuration state machine may allow a local attacker t |
| CVE-2025-69248 | 1 | no | free5GC has Array Index Out of Bounds in AMF Leading to Denial of Service |
| CVE-2021-41773 | 1 | Yes | Path traversal and file disclosure vulnerability in Apache HTTP Server 2.4.49 |
| CVE-2024-37018 | 1 | no | The OpenDaylight 0.15.3 controller allows topology poisoning via API requests because an application can manip |
Bug bounties
# HackerOne: publicly disclosed reports per year
12 759 reports that programmes chose to disclose publicly. Disclosed reports are a small, self-selected share of all reports.
- Disclosed reports
- With a bounty
# Bounties on disclosed reports (USD)
2 275 disclosed reports show a bounty; the median was $500, the largest $50 000.
# Programmes with the most disclosed reports
# Weaknesses in disclosed reports
# Most disclosed hackers
Reporters by number of publicly disclosed reports.