The CWE Top 25 Most Dangerous Software Weaknesses is an annual, data-driven ranking published by The MITRE Corporation under sponsorship from CISA, as part of the Common Weakness Enumeration (CWE) program; the current, 2025 edition tops the list with CWE-79 Cross-Site Scripting (score 60.38), computed from 39,080 CVE Records published between 1 June 2024 and 1 June 2025. MITRE's own description frames the purpose directly: these weaknesses are "often easy to find and exploit" and "can lead to exploitable vulnerabilities that allow adversaries to completely take over a system, steal data, or prevent applications from working." Unlike a vulnerability database entry for one specific flaw, each CWE in the list is a class of weakness — the root cause pattern behind many individual CVEs — which is what makes the list useful for prioritizing where secure-coding effort and review time go, rather than reacting one CVE at a time.
Key takeaways
- Frequency Fr(CWE_X) — how often that CWE is mapped to a CVE Record in the NVD dataset for the analysis window, normalized between the minimum and maximum frequency observed across all CWEs.
- Per MITRE's 2025 methodology page, the current list analyzed 39,080 CVE Records published between 1 June 2024 and 1 June 2025.
- Source: MITRE, 2025 CWE Top 25, read 10 November 2026.
- MITRE's 2025 key insights name the biggest movers: missing authorization (CWE-862) rose from #9 to #4, NULL pointer dereference (CWE-476) from #21 to #13, and missing authentication (CWE-306) from #25 to #21.
How is the score calculated?
MITRE's published methodology defines the danger score for each CWE as:
Score(CWE_X) = Fr(CWE_X) × Sv(CWE_X) × 100
- Frequency
Fr(CWE_X)— how often that CWE is mapped to a CVE Record in the NVD dataset for the analysis window, normalized between the minimum and maximum frequency observed across all CWEs. - Severity
Sv(CWE_X)— the average CVSS base score (v3.0/v3.1 only) across CVEs mapped to that CWE, normalized the same way.
MITRE states this multiplication is deliberate: "weaknesses that are both common and caused significant harm will receive the highest scores," while a weakness whose exploitation "was of low impact will not receive a high Severity score, regardless of how common it was in the dataset." A weakness has to be both common and severe to rank highly — high frequency alone, or high severity alone, isn't enough.
What did the 2025 analysis cover?
Per MITRE's 2025 methodology page, the current list analyzed 39,080 CVE Records published between 1 June 2024 and 1 June 2025. Of those, 9,468 records (24%) were flagged for expert remapping, and 2,459 (26% of the flagged set) received review from CVE Numbering Authorities (CNAs) before final scores were computed. MITRE documents two methodology changes introduced for 2025: CWE mappings were used "as provided, without normalizing them back to View-1003" (a simplified mapping view used in prior years), and an LLM-based tool assisted with mapping suggestions during remapping for the first time. Both changes mean 2025 scores and rankings are not perfectly comparable, year over year, to the 2024 list computed under the prior methodology.
What is on the full 2025 list?
The 2025 list page gives all 25 entries with a score, the number of their CVEs that are in CISA's Known Exploited Vulnerabilities (KEV) catalog, and the rank change against 2024 ("new" = no previous rank):
| Rank | CWE-ID | Name | Score | CVEs in KEV | Change |
|---|---|---|---|---|---|
| 1 | CWE-79 | Cross-Site Scripting | 60.38 | 7 | 0 |
| 2 | CWE-89 | SQL Injection | 28.72 | 4 | +1 |
| 3 | CWE-352 | Cross-Site Request Forgery (CSRF) | 13.64 | 0 | +1 |
| 4 | CWE-862 | Missing Authorization | 13.28 | 0 | +5 |
| 5 | CWE-787 | Out-of-bounds Write | 12.68 | 12 | −3 |
| 6 | CWE-22 | Path Traversal | 8.99 | 10 | −1 |
| 7 | CWE-416 | Use After Free | 8.47 | 14 | +1 |
| 8 | CWE-125 | Out-of-bounds Read | 7.88 | 3 | −2 |
| 9 | CWE-78 | OS Command Injection | 7.85 | 20 | −2 |
| 10 | CWE-94 | Code Injection | 7.57 | 7 | +1 |
| 11 | CWE-120 | Classic Buffer Overflow | 6.96 | 0 | new |
| 12 | CWE-434 | Unrestricted Upload of File with Dangerous Type | 6.87 | 4 | −2 |
| 13 | CWE-476 | NULL Pointer Dereference | 6.41 | 0 | +8 |
| 14 | CWE-121 | Stack-based Buffer Overflow | 5.75 | 4 | new |
| 15 | CWE-502 | Deserialization of Untrusted Data | 5.23 | 11 | +1 |
| 16 | CWE-122 | Heap-based Buffer Overflow | 5.21 | 6 | new |
| 17 | CWE-863 | Incorrect Authorization | 4.14 | 4 | +1 |
| 18 | CWE-20 | Improper Input Validation | 4.09 | 2 | −6 |
| 19 | CWE-284 | Improper Access Control | 4.07 | 1 | new |
| 20 | CWE-200 | Exposure of Sensitive Information | 4.01 | 1 | −3 |
| 21 | CWE-306 | Missing Authentication for Critical Function | 3.47 | 11 | +4 |
| 22 | CWE-918 | Server-Side Request Forgery (SSRF) | 3.36 | 0 | −3 |
| 23 | CWE-77 | Command Injection | 3.15 | 2 | −10 |
| 24 | CWE-639 | Authorization Bypass Through User-Controlled Key | 2.62 | 0 | +6 |
| 25 | CWE-770 | Allocation of Resources Without Limits or Throttling | 2.54 | 0 | +1 |
Source: MITRE, 2025 CWE Top 25, read 10 November 2026. The KEV column tells a different story from the score: OS command injection (CWE-78) ranks 9th by score but has the most KEV entries (20), ahead of use after free (14), out-of-bounds write (12), deserialization (11) and missing authentication (11). Treat any specific rank or score as current as of the 2025 edition rather than permanent; MITRE re-runs the analysis annually against a fresh CVE window.
What moved, and what dropped off in 2025?
MITRE's 2025 key insights name the biggest movers: missing authorization (CWE-862) rose from #9 to #4, NULL pointer dereference (CWE-476) from #21 to #13, and missing authentication (CWE-306) from #25 to #21. Three buffer overflow children (CWE-120, CWE-121, CWE-122) and improper access control (CWE-284) entered for the first time. Six entries fell off:
| CWE | Name | 2024 rank | 2025 rank |
|---|---|---|---|
| CWE-269 | Improper Privilege Management | 15 | 29 |
| CWE-190 | Integer Overflow or Wraparound | 23 | 30 |
| CWE-287 | Improper Authentication | 14 | 31 |
| CWE-400 | Uncontrolled Resource Consumption | 24 | 32 |
| CWE-798 | Use of Hard-coded Credentials | 22 | 35 |
| CWE-119 | Improper Restriction of Operations within the Bounds of a Memory Buffer | 20 | 39 |
MITRE explains most of these drops by the end of View-1003 normalization, a set of 130 weaknesses NVD uses for enrichment. In earlier years, mappings to a child CWE were rolled up to its parent; in 2025 they stayed with the child. CWE-119 fell because its children CWE-120, 121 and 122 now count separately. MITRE gives CWE-269 as an example: it would have had 633 CVEs with its children rolled up, enough to stay in the list. For a device maker, the drop of hard-coded credentials (CWE-798) to #35 says more about mapping practice across all software than about firmware, where default and embedded credentials remain a familiar finding.
Which Top 25 weaknesses matter most in embedded firmware?
For connected devices, the list's web-application entries (XSS, SQL injection, SSRF) still apply to the admin UI, but the entries with the most KEV weight sit in code that every router, camera, gateway and controller has: a web server, a C protocol parser and a management interface. CISA and FBI made the same point in their July 2024 Secure by Design Alert on OS command injection, which cites three exploited network edge device CVEs (CVE-2024-20399, CVE-2024-3400, CVE-2024-21887) and asks manufacturers to generate commands with functions that keep data separate from command syntax.
| Weakness group (2025 Top 25) | Typical place in device firmware | Fix guide |
|---|---|---|
| CWE-78, CWE-77 command injection | Web UI diagnostics (ping, traceroute, NTP) that build a shell string | Command injection |
| CWE-22 path traversal, CWE-434 dangerous upload | File download, log export, configuration and firmware upload handlers | Path traversal |
| CWE-787, 125, 120, 121, 122, 416, 476 memory safety | Protocol parsers (HTTP, UPnP, DHCP, Modbus), string handling in C | Memory corruption |
| CWE-306, 862, 863, 284, 639 missing or broken access control | Management APIs, debug services, cloud pairing endpoints | Authentication bypass |
| CWE-798 hard-coded credentials (#35 in 2025) | Default accounts, embedded keys, service passwords | Hard-coded credentials |
Is there a hardware version of the list?
Yes. MITRE publishes a separate CWE Most Important Hardware Weaknesses list. The 2025 edition updates the October 2021 list. It is built differently from the software list: per MITRE, it combines AI-assisted data collection with expert opinion from the Hardware CWE Special Interest Group rather than the CVE-based score. Chip and module makers, and device makers that own their board design, should read both lists.
How should a device maker use the list?
The list is most useful as input to secure development and to PSIRT root-cause work, not as a scan target. A practical loop:
Tag every vulnerability you fix with a CWE, in your tracker and in the advisory. Use the most specific CWE that fits; MITRE marks each entry with a mapping usage of Allowed, Allowed-with-Review, Discouraged or Prohibited. Count your own CWEsover the last two or three years of advisories and internal findings. Compare with the Top 25 and its KEV columnClasses that are both frequent in your code and heavily exploited across the industry come first. Turn each priority class into controls: a static analysis rule, a banned-function list (for example system()with user input,strcpy), a code review checklist item, and a fuzzing target for the parser concerned. IEC 62443-4-1 asks for this kind of secure implementation practice; see our IEC 62443-4-1 guide.Test for the classesin your firmware security testing, following the OWASP firmware testing methodology. Re-run the count once a yearwhen MITRE publishes a new list, and check whether your own counts drop.
Worked example: what does root-cause tracking with the list look like?
Synthetic example. A made-up maker of industrial gateways tallies the 12 advisories it published over two years:
| CWE | Own advisories | 2025 Top 25 rank | CVEs in KEV (2025 list) | Decision |
|---|---|---|---|---|
| CWE-78 OS command injection | 4 | 9 | 20 | Priority 1: replace shell calls in the web UI, add a review rule |
| CWE-787 out-of-bounds write | 3 | 5 | 12 | Priority 2: fuzz the Modbus and HTTP parsers in CI |
| CWE-798 hard-coded credentials | 2 | 35 (dropped) | — | Keep: own data outweighs the rank; per-device credentials at manufacturing |
| CWE-306 missing authentication | 2 | 21 | 11 | Priority 3: authentication on every management endpoint by default |
| CWE-22 path traversal | 1 | 6 | 10 | Shared library for path normalization |
Two points stand out. CWE-798 left the Top 25, but it is a sixth of this maker's own history, so it stays on the list of controls. And CWE-78 outranks CWE-787 here because both the maker's own data and the KEV column point to it, even though the score order is the reverse.
How do you avoid over-reading the list?
The Top 25 ranks classes of weakness, not specific products or vendors, and the ranking reflects what was reported and scored in the public CVE dataset during the analysis window — a weakness that's genuinely common in embedded or industrial code but under-reported to CVE (because fewer researchers look at that code, or fewer vendors file CVEs for it) can rank lower than its real-world prevalence would suggest. Teams should weigh it against their own CVE history and against vulnerability-specific prioritization data (see our KEV + EPSS prioritization guide and CVSS v4 vs v3.1 guide) rather than treating rank order alone as a remediation queue. Common pitfalls:
- Comparing scores across years. The 2025 methodology changed; a rank change can be a mapping effect, as with CWE-119.
- Ignoring the KEV column. Score order and exploitation order differ, as CWE-78 shows.
- Mapping to abstract CWEs. CWE-20 (improper input validation) is too broad to guide a fix; pick the specific child.
- Using the list as a compliance checklist. No regulation requires "Top 25 coverage"; standards ask for a secure development process.
- Treating a clean component scan as proof. A known-CVE scan of third-party code says nothing about weaknesses in your own code.
Where does SoftWSP fit, and where does it not?
SoftWSP covers the known-vulnerability side: it extracts an SBOM from a firmware image, matches the components against NVD, CISA KEV, FIRST EPSS and CISA CSAF ICS advisories, and drafts a cited report that an engineer reviews before it ships. Each matched CVE links to its public record, where the CVE's CWE assignment can be read. SoftWSP does not analyze your own source code or binaries for new weakness instances, does not group findings by CWE class, and does not do reachability analysis; use static analysis, code review and fuzzing for that. To see what a component inventory looks like, try the SBOM viewer or read the firmware SBOM guide; for a scan of a specific image, see device & firmware security services.
This guide summarises a public MITRE/CISA-sponsored CWE program publication; it does not constitute legal or compliance advice.
Put this guide to work
Frequently asked questions
What is the CWE Top 25?
An annual, data-driven ranking of the most dangerous software weaknesses, published by The MITRE Corporation under sponsorship from the US Department of Homeland Security's Cybersecurity and Infrastructure Security Agency (CISA), as part of the Common Weakness Enumeration (CWE) program. MITRE's own description of the Top 25 states these weaknesses are "often easy to find and exploit" and "can lead to exploitable vulnerabilities that allow adversaries to completely take over a system, steal data, or prevent applications from working," and that the list is meant to guide "investments, policies, and practices" by highlighting root causes rather than individual vulnerabilities.
How is the Top 25 score calculated?
Per MITRE's own published methodology, each CWE's danger score is Score(CWE_X) = Fr(CWE_X) × Sv(CWE_X) × 100, where Frequency Fr(CWE_X) is the count of CVE Records mapped to that CWE in the National Vulnerability Database (NVD), normalized against the minimum and maximum counts observed across all CWEs in the dataset, and Severity Sv(CWE_X) is each CWE's average CVSS base score (v3.0/v3.1 only), normalized the same way. MITRE states this combination means "weaknesses that are both common and caused significant harm will receive the highest scores," while a low-impact weakness "will not receive a high Severity score, regardless of how common it was in the dataset" — frequency alone cannot push a weakness up the list.
What data does the current (2025) list analyze, and what changed in its methodology?
MITRE's 2025 methodology page states the analysis covered 39,080 CVE Records published between 1 June 2024 and 1 June 2025, of which 9,468 (24%) were flagged for expert remapping and 2,459 (26% of those) received CNA (CVE Numbering Authority) review before scores were finalized. MITRE states two methodology changes for 2025: for the first time, the Top 25 used CWE mappings "as provided, without normalizing them back to View-1003" (a simplified subset view used in earlier years), and an LLM-based tool was used for the first time to assist with mapping suggestions during the remapping process.
What are the top entries on the current CWE Top 25 list?
Per MITRE's 2025 list, the top five are: #1 CWE-79 Cross-Site Scripting (score 60.38), #2 CWE-89 SQL Injection (28.72), #3 CWE-352 Cross-Site Request Forgery (13.64), #4 CWE-862 Missing Authorization (13.28), and #5 CWE-787 Out-of-bounds Write (12.68). Rankings and exact scores shift from year to year as the underlying CVE dataset and remapping change — teams tracking a specific weakness should check MITRE's current list rather than treat any single year's rank as fixed.
Sources
- MITRE: CWE Top 25 Most Dangerous Software Weaknesses — current list landing page
- MITRE: 2025 CWE Top 25 Most Dangerous Software Weaknesses — sponsor, purpose, list
- MITRE: 2025 CWE Top 25 Methodology — scoring formula, dataset, remapping, LLM-assisted mapping change
- MITRE: 2024 CWE Top 25 Methodology (prior-year formula and dataset, for comparison)
- NIST: National Vulnerability Database (NVD) — official site
- MITRE: 2025 CWE Top 25 Key Insights (movers, new entries, drop-outs, View-1003)
- MITRE: CWE-1003 Weaknesses for Simplified Mapping of Published Vulnerabilities
- MITRE: CWE Most Important Hardware Weaknesses (2025 update)
- CISA and FBI: Secure by Design Alert: Eliminating OS Command Injection Vulnerabilities (10 July 2024)
- FIRST: CVSS v3.1 Specification Document
About this guide
SoftWSP Firmware Security Desk
Covers the Firmware vulnerabilities & fixes cluster: 131 guides.

In this cluster
Questions about this page?
