Ruby2Shell: MITRE says 7.4 High, Ruby's advisory still says no CVE assigned, Veria says default Rails is affected
Read on 20 November 2026: one of Ruby's seven flaws now has a CVE id, CVE-2026-132825 at 7.4 High, while the maintainers' advisory still says none are assigned. Veria says default Rails and classic Sinatra are reachable; a watchTowr specialist says production exploitation is challenging.
By Parminder Kumar Sharma · LinkedIn · · 25 min read
One of Ruby's seven flaws has a CVE id since 19 November; the maintainers' page, read on 20 November, still lists none
Ruby's maintainers dated their advisory on seven flaws 18 November 2026, and brief 746 read it the next day. It named every flaw by GitHub Security Advisory (GHSA) id only, and said CVE identifiers "have been requested for these vulnerabilities but have not been assigned yet". This site read the same page again at about 15:19 UTC on 20 November. It still says that, and its text is the same as the copy this site saved at about 12:16 UTC on 19 November, 27 h 03 min earlier (derived; a line by line comparison of the two saved files found only the header line that records when each was saved different).
In the meantime CVE.org has gained a record. CVE-2026-132825 was assigned by mitre and published at 06:34:26 UTC on 19 November, 32 h 44 min before this site's re-read (derived). Its description reads: "Ruby before 3.3.14, 3.4.13, and 4.0.10 has a heap-based buffer overflow during case conversion." It scores 7.4 High on CVSS 3.1 with a vector that starts AV:L, local. The description does not name a GHSA id, but the record's reference list includes the GitHub advisory page for GHSA-34m4-m9fh-gqp5, the one byte heap overflow in case conversion that Ruby's advisory lists first. So one of the seven flaws has an id on the saved pages and six do not (derived: 7 minus 1).
That is the checkable fact. Four other pages describe the flaw, under the name Ruby2Shell that Veria Labs gave it. They disagree about how bad it is and who can reach it, and each disagreement is printed below. Here is what the fact does not establish:
- That Ruby has been told about the id. The maintainers' page, as saved, does not mention CVE-2026-132825. It may be updated at any time. This site did not read the GitHub advisory page, the release posts or Ruby's security index on 20 November.
- That the other six flaws have no CVE ids. The saved pages show none. This site did not search the CVE list for other ids, so it cannot say none exist.
- That anyone is exploiting the flaw. The CISA rating inside the record says Exploitation none at 18:55 UTC on 19 November; Previdian and watchTowr each say none reported or confirmed, as of 19 November. Nothing saved is newer than those statements.
- That a default Rails application is exposed. Veria says so. The Ruby advisory and the record print no framework, and a watchTowr specialist, quoted by SecurityWeek, calls production exploitation likely to be challenging.
- A public proof of concept. Veria's page, saved at about 15:18 UTC on 20 November, says it will publish one on 25 November 2026 or later. It had not been published on that page.
What the CVE record says, and where it points
The record is short. Its author is mitre, through the CVE Request form (the record's generator field says "CVE-Request-form"). It is a CVE JSON 5.2 record with one affected product, Ruby by ruby-lang, a default status of unaffected, and three version ranges. Table 1 lists the fields as printed.
Table 1. CVE-2026-132825, fields as the record prints them (CVE.org JSON record, saved 20 November 2026).
| Field | As printed | Note |
|---|---|---|
| Assigner, state | mitre, PUBLISHED | The assigner is not Ruby. The record carries no credits and no fix field. |
| Reserved, published | 19 Nov 2026, 06:34:25 and 06:34:26 UTC | Reserved and published within one second (derived from the two stamps). |
| Last updated | 20 Nov 2026, 04:06:08 UTC | The mitre container itself still carries its publication stamp; the later stamp comes from the CVE Program container. |
| Description | Ruby before 3.3.14, 3.4.13, and 4.0.10 has a heap-based buffer overflow during case conversion. One byte can be written outside a buffer for, for example, ISO-8859-1 or Windows-1252. | It names no GHSA id and no framework. |
| Weakness | CWE-122 Heap-based Buffer Overflow | |
| CVSS 3.1 | 7.4, HIGH. CVSS:3.1/AV:L/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H | Local, high complexity, no privileges, no user interaction. |
| Affected ranges | Version 0 to before 3.3.14; 3.4.0 to before 3.4.13; 4.0.0 to before 4.0.10 | The first range has no lower bound printed. |
| References, mitre container | Three: Ruby's advisory page, the GitHub advisory page for GHSA-34m4-m9fh-gqp5, and a discussion thread | This is where the record links the id to the GHSA entry. |
| CISA-ADP rating | Exploitation none; Automatable no; Technical Impact total, stamped 19 Nov 2026, 18:55:33 UTC | A rating at that time, not a statement about now. |
The record's references include Ruby's own advisory, so mitre points readers at the page that says no ids are assigned. The CVE Program container, stamped 04:06 UTC on 20 November, lists seven further references: an Ubuntu security page, a SUSE bug tracker entry, Veria's page, a pull request in a Semgrep detection scripts repository, a commit in the omnibus GitLab repository, a post on the oss-security mailing list, and one further code repository on GitHub. This site did not open any of them. It does not name the code repository, because it did not read it and cannot say what it holds.
The link between the CVE and the GHSA entry therefore exists in three places that this site read: the record's reference list, Veria's page header, and Previdian's table, which prints "CVE-2026-132825" and "GHSA-34m4-m9fh-gqp5" side by side. It exists on no page of Ruby's that this site read.
The clock: 6 h 34 min from Ruby's footer to the record, 32 h 44 min to this site's re-read
Diagram 1 puts every time stamp on one axis, in hours. The Ruby page is the awkward one. The post is headed "Posted by" with the date 18 Nov 2026 and no time, while its History line says "Originally published at 2026-11-19 00:00:00 (UTC)". If the footer is the time, the CVE record came 6 h 34 min after it (derived: 00:00:00 to 06:34:26). If the post went up at some point on 18 November, the gap is anywhere from 6 h 34 min to 30 h 34 min (derived). The saved pages print no other time for the post, so this brief gives no single gap for it. The pages do not say why the footer shows midnight.
Table 2. Gaps from the CVE record (published 19 November 2026, 06:34:26 UTC) to each later stamp, derived from the saved pages.
| Stamp | Time (UTC) | After the record |
|---|---|---|
| Veria page, datePublished in its metadata | 19 Nov, 03:45:00 | Before it: the record came 2 h 49 min later |
| Previdian, datePublished in its metadata | 19 Nov, 13:13:44 | 6 h 39 min |
| CISA-ADP rating inside the record | 19 Nov, 18:55:33 | 12 h 21 min |
| watchTowr, published | 19 Nov, 20:54:54 | 14 h 20 min |
| Record last updated | 20 Nov, 04:06:08 | 21 h 31 min |
| SecurityWeek, published | 20 Nov, 10:34:10 | 27 h 59 min |
| This site re-reads the Ruby advisory | 20 Nov, about 15:19 | 32 h 44 min |
Two readings of the order follow, and this site does not choose. Veria's metadata puts its page before the record by 2 h 49 min, yet the record's own list of references links to Veria's page, and Veria's page prints the CVE id. So either the page was edited to add the id after the record appeared (its visible Updated line is 20 November, 15:00 UTC), or the metadata date is not the date the id was added. The saved pages do not say. Brief 746's own copy of the Ruby page was saved 5 h 41 min after the record was published (derived), so that copy was taken after the record existed, and the advisory text it holds is the same as today's.
7.4 High on the record, Critical on two pages, and a vector that says local
Three severities are printed, by three different parties. The record, assigned by mitre, scores 7.4 High on CVSS 3.1. Veria's page header says "Severity Critical" and prints no number or vector. Previdian prints "Critical (Ruby advisory); 7.4 High (MITRE CVSS v3.1)", attributing Critical to the Ruby advisory. The saved Ruby advisory prints no severity word and no score: this site searched its text for Critical, High, CVSS and severity and found none. This site did not read the GitHub advisory page, which may carry a rating, so it cannot check what Previdian means. watchTowr and SecurityWeek both print 7.4.
The vector letters are the part a reader can check. This site recomputed the score from the record's vector with the CVSS 3.1 formula and got 7.4 (derived), so the number follows from the letters. The letters say local (AV:L), high complexity (AC:H), no privileges (PR:N) and no user interaction (UI:N), with high impact on confidentiality, integrity and availability. Veria's header says "Access needed" is "Network, no authentication". Two of those agree: no privileges and no authentication. The third does not: local against network. Previdian prints the mismatch itself, and calls the record's vector "MITRE's initial".
Two what-if calculations show how much rests on those letters. Changing only the attack vector to network gives 8.1 under the same formula (derived), still High, 0.7 higher. Changing the attack vector to network and the complexity to low gives 9.8 (derived), which is Critical. Under the CVSS 3.1 rating bands (7.0 to 8.9 High, 9.0 to 10.0 Critical, from the specification, not a saved page) the label Critical needs both changes. These are this site's arithmetic. They are not scores anyone assigned, and no saved page says why Veria uses Critical.
Table 3. The severity statements, by who prints them.
| Party | What it prints | What it does not print |
|---|---|---|
| CVE record (mitre) | 7.4 High, CVSS 3.1, AV:L/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H | A rationale; any framework |
| Veria Labs | "Severity Critical"; Access needed "Network, no authentication" | A score, a vector or a scale |
| Previdian | Critical (attributed to the Ruby advisory); 7.4 High (MITRE CVSS v3.1) | Where in the Ruby advisory the word is |
| Ruby advisory (saved copy) | No severity word, no score | Any CVSS vector |
| watchTowr, SecurityWeek | 7.4 (v3.1) and 7.4 | A separate assessment |
Reach: four pages, four wordings
Diagram 2 sets the six pages side by side on id, score, reach, difficulty and exploitation. The reach column is where they differ most.
Ruby's advisory prints a condition for its four pattern compilation flaws, applications that compile attacker controlled patterns, and none for case conversion: the entry says only that "Changing the case of a crafted string may allow remote code execution". The CVE record says one byte can be written outside a buffer "for, for example, ISO-8859-1 or Windows-1252". The words "for example" make those two encodings a sample. Previdian says legacy single byte encodings "including" the two. Veria's page says the flaw is in several ISO-8859 and Windows-125x encodings and is reached through four operations, upcase, swapcase, downcase(:fold) and casecmp?, and that UTF-8 strings take a different code path and are not affected. Previdian prints the same four names and the same UTF-8 statement.
On who can reach it, Veria's framework text says the bug "is reachable from the following frameworks in their default configuration", and names Rails and classic style Sinatra. It says other frameworks and plain Rack are "not exposed by default", and adds that application code that upcases strings received in a legacy encoding may also be vulnerable. The word unauthenticated is not in that framework text. It is in Veria's page header ("Network, no authentication") and in the page's metadata description, which says the flaw is "reachable without authentication in common Ruby web frameworks and self-hosted software". Previdian repeats it for default Rails and classic Sinatra. SecurityWeek writes that unauthenticated attackers could trigger the defect "from Rails, classic-style Sinatra, and potentially other frameworks", where Veria's text says other frameworks are not exposed by default: a difference between a secondary and its source.
watchTowr goes the other way. It lists two conditions: the application must process text in a single byte encoding such as ISO-8859-1 or Windows-1252 "rather than Ruby's default UTF-8", and it must run a case changing function on text an attacker controls. It also says Rails and classic Sinatra applications "are affected in their default configuration", along with "any Ruby application that changes the case of external text in a single-byte encoding". Read together, the pages say two things that sit awkwardly: that the flaw needs text in a single byte encoding, and that default Rails is affected. No saved page says what in a default Rails application puts external text into a single byte encoding, or what makes a configuration non default. This site does not guess, and prints nothing about how an attacker would supply the text.
How hard: a private proof of concept, a promised public one, and a specialist who says production is challenging
The Ruby advisory says "We are aware of PoCs that prove this flaw can be used to trigger RCE in some scenarios" about this flaw. Veria says it "found this vulnerability independently and was the first to demonstrate remote code execution". Previdian writes "Private PoCs acknowledged". None of the three says who holds a proof of concept other than Veria, or what the scenarios are.
Veria's date is the checkable item. Its page says it will publish a proof of concept "demonstrating remote code execution on stock Rails" on 25 November 2026, "or later if major vendors have not shipped a fix by then". That is 7 days after the 18 November post date and 5 days after the 20 November reading (derived). Previdian says Veria plans to publish "approximately one week after disclosure". The page does not name the major vendors. Of the six products in Veria's own table, one, GitHub Enterprise Server, is marked Vulnerable with no fixed release (read on 20 November); whether that product is one of the vendors the sentence means is this site's inference, and the page does not say so.
The other view comes from SecurityWeek, a trade news site and a secondary source, quoting a watchTowr specialist whom this brief does not name. It says successful exploitation in production "is likely to be challenging" because laying out memory correctly is difficult, because an attacker would need the right non-glibc memory allocator, runtime configuration and application logic preconditions, because other people on the internet would be attacking the same system, and because address space layout randomisation would have to be defeated. It adds: "Most likely, malformed requests could lead to denial-of-service (DoS) and service disruption". watchTowr's own FAQ says reaching code execution "takes further work specific to each application".
These two views are not the same claim. Veria says it has shown remote code execution on a stock Rails application. watchTowr's specialist says production exploitation is challenging. Both can be true: a demonstration on a prepared target and a difficult attack on a varied estate are different things, but that is this site's reading and neither page says it. The record's own vector letter for complexity, AC:H, high, agrees in direction with the specialist and not with Veria's header.
Versions and the end of life line
Ruby's advisory, the record, Veria, Previdian and watchTowr agree on the three fixed releases: 4.0.10, 3.4.13 and 3.3.14. The affected ranges agree in substance across Ruby, the record and Veria. They differ in what they print about older releases. Diagram 3 shows the grid.
Ruby says "Ruby 3.2 and earlier have reached their end of life and will not receive these fixes". Veria's table prints a row for "3.2 and older" with "All releases back to 2.4.0" in the affected column and "EOL" in the fixed column, and Previdian's table prints "End of life, no fix" for the releases from 2.4.0 to 3.2.x. The CVE record prints its first range as version 0 to before 3.3.14, with no lower bound; Ruby's advisory lists no range below 3.3. So the record's wording would put every release below 3.3.14 inside the affected set, including the end of life ones, while the maintainers' page lists none of them as affected and says they get no fix. This site reads that as a difference in how the pages mark the range, not as evidence that 3.2 is safe: Veria and Previdian both say it is affected back to 2.4.0.
How wide is the unfixed span? Fixes exist for three release series: 4.0, 3.4 and 3.3. Veria's range for 3.2 and older starts at 2.4.0. With the usual major dot minor numbering, that spans seven series (2.4, 2.5, 2.6, 2.7, 3.0, 3.1 and 3.2), so seven series with no fix against three with one (derived; it assumes that numbering, which no saved page lists). Veria says the affected code "has been there since Ruby 2.4.0". This brief gives no release dates for any of them. None is printed on the saved pages and this site does not supply them from memory.
watchTowr's version table prints Ruby 3.2 and earlier as "Not yet available" in its fixed column, and its text says "no fix exists for it" and that those systems should move to a supported release line. The first reads as if a fix may come. Ruby's page says it will not.
Products that bundle their own Ruby: five patched, one vulnerable, and two statuses that changed in a day
Veria's page lists six self hosted products that bundle their own Ruby. It says that for those the fix "has to come from the vendor", and that products marked Vulnerable "have no fixed release yet: every current version is vulnerable". Table 4 copies the table as printed, with the status words.
Table 4. Veria Labs's self hosted software table as printed. The page says Updated 20 November 2026 at 15:00 UTC; as saved at about 15:18 UTC.
| Software | Affected versions | Fixed in | Status |
|---|---|---|---|
| GitHub Enterprise Server | 3.22.5 and earlier | None | Vulnerable |
| Chatwoot | 4.20.0 and earlier | 4.20.1 | Patched |
| GitLab | 19.6.0 and earlier | 19.6.1, 19.5.8, 19.4.10 | Patched |
| Mastodon | 4.7.5 and earlier | 4.7.6, 4.6.12, 4.5.22, 4.4.29 | Patched |
| OpenProject | 17.9.7 and earlier | 17.9.8, 17.8.3 | Patched |
| Discourse base image | Before 2.0.20261119-0203 | 2.0.20261119-0203 | Patched |
Counting the table, five of six are marked Patched and one, GitHub Enterprise Server 3.22.5 and earlier, is marked Vulnerable with no fixed release (derived). That is an account of what Veria's page says on 20 November. The vendors' own pages are not among the saved sources, and this brief does not verify any status against them.
The same list was different a day earlier. Previdian, in a page published on 19 November (13:13 UTC per its metadata), says Veria then listed GitLab 19.6.0, GitHub Enterprise Server 3.22.5 and Chatwoot 4.20.0 as affected, and Mastodon 4.7.6, OpenProject 17.9.8 and the standard Discourse installation as fixed: three and three (derived count). Veria's table as saved on 20 November lists GitLab and Chatwoot as Patched, with fixed versions 19.6.1, 19.5.8 and 19.4.10 for GitLab and 4.20.1 for Chatwoot. SecurityWeek, published at 10:34 UTC on 20 November, lists the same five products as patched and GitHub Enterprise Server as affected, which agrees with Veria's table. So two statuses moved from affected to patched between Previdian's account and Veria's table, and Veria's statuses are the only account of it this site saved.
A status word is also the vendor's claim, not a test of your installation. Previdian itself warns that "Deployment-specific verification remains necessary", and its page says updating the system Ruby alone "may not be sufficient" for an application that bundles its own runtime.
Three labels doing the work of a control
"Not exposed by default" is the label that most readers will want to read as safe. Veria uses it for other frameworks and plain Rack, and in the same sentence says application code that upcases strings in a legacy encoding "may also be vulnerable". A default is a configuration, and the pages do not define what makes yours non default. The label tells a reader where Veria looked, not what is true of the application in front of them. A team that runs plain Rack and reads this as an all clear has not checked whether any of its own code changes the case of outside text.
"Critical" and "7.4 High" are labels on two different things. One is a headline word with no scale printed beside it. The other is a number that follows from a vector the same record prints as local. Sorting a queue by either decides how fast a patch goes out, and the two sort the same flaw into different places.
"Patched" in Veria's product table is a status word for a vendor release, not for a running system. The pages say to verify the runtime that is actually running, rebuild images and restart. A container image that predates the fix can sit on a host with a patched Ruby installed beside it.
Where the pages differ, with nothing resolved
Table 5. Differences between and inside the saved pages. Each is reported as printed; none is resolved.
| Point | One page prints | Another prints |
|---|---|---|
| CVE ids | Ruby advisory, re-read 20 Nov: requested, not assigned, GHSA ids only | CVE record, 19 Nov 06:34 UTC: CVE-2026-132825. No saved page shows an id for the other six flaws |
| Severity | Record: 7.4 High (CVSS 3.1, AV:L). Veria: "Critical", no score | Previdian: "Critical (Ruby advisory)", though the saved Ruby page prints no severity word |
| Attack vector | Record vector: AV:L, local | Veria header: "Network, no authentication" |
| Reach | Veria: Rails and classic Sinatra; other frameworks and plain Rack not exposed by default | watchTowr: also "any Ruby application that changes the case of external text in a single-byte encoding". SecurityWeek: "potentially other frameworks" |
| Difficulty | Veria and Previdian: remote code execution shown or acknowledged, public PoC planned | A watchTowr specialist, via SecurityWeek: production exploitation likely challenging, denial of service more likely |
| Encodings | Record: ISO-8859-1 and Windows-1252 "for, for example" | Previdian: "including" the two. Veria: several ISO-8859 and Windows-125x encodings. watchTowr: "such as" the two |
| 3.2 and older | Ruby: end of life, will not receive these fixes. Veria and Previdian: affected back to 2.4.0 | Record: version 0 to before 3.3.14. watchTowr table: fixed version "Not yet available" |
| Product statuses | Previdian, 19 Nov: GitLab and Chatwoot affected | Veria table, 20 Nov 15:00 UTC: both Patched |
| Veria PoC timing | Veria: 25 November 2026, or later if major vendors have not shipped a fix | Previdian: "approximately one week after disclosure" |
| Date of the Ruby post | Page header: 18 Nov 2026 | History line: 2026-11-19 00:00:00 (UTC) |
What the pages state, and what they do not
Table 6. What the saved pages establish and what they leave open.
| Question | Stated | Not stated |
|---|---|---|
| Which Ruby releases fix it? | Ruby, record, Veria, Previdian, watchTowr: 4.0.10, 3.4.13, 3.3.14 | Any fix for 3.2 and older; any release dates |
| Does the flaw have a CVE id? | One: CVE-2026-132825, mitre, 19 Nov 06:34 UTC, linked to GHSA-34m4-m9fh-gqp5 in the record's references | Ids for the other six; a Ruby page that mentions the id |
| How severe is it? | 7.4 High (record, CVSS 3.1); Critical (Veria, Previdian) | A scale or vector for Critical; why the vector is local |
| Who can reach it? | Veria: default Rails and classic Sinatra. watchTowr: also any app that changes the case of outside text in a single byte encoding | What makes a configuration non default; how many applications read single byte text |
| Is it exploited? | CISA-ADP rating Exploitation none (19 Nov 18:55 UTC); Previdian and watchTowr: none confirmed or reported as of 19 Nov | Anything after those stamps; any catalogue entry (none saved) |
| Is a public PoC out? | Not on Veria's page as saved; planned for 25 Nov or later | Whether another party has published one; what the CVE Program container's further GitHub reference holds |
| What is the workaround? | Upgrade. Previdian: restricting public access can reduce exposure | A vendor confirmed workaround (watchTowr: none published) |
| Who is credited? | The Ruby advisory credits individuals and organisations per flaw; Veria says it found this one independently | This brief prints no individual's name |
The record, the version tables and the product list, as printed
This section is defensive. The saved pages print no indicators of compromise, no log lines, no detection rule and no reports of exploitation, and this site prints none. Veria's page also carries a technical section with code from the flawed routine and the byte level detail of the overflow. This brief does not reproduce it, and prints no input, no sequence and no byte values. What follows is the set of items a defender can use to check versions and scope the review of application code.
The first block is the CVE record's own text and numbers, copied from the JSON. It is the only place the vector letters appear.
description: Ruby before 3.3.14, 3.4.13, and 4.0.10 has a heap-based buffer overflow during case conversion. One byte can be written outside a buffer for, for example, ISO-8859-1 or Windows-1252.
vectorString: CVSS:3.1/AV:L/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H
baseScore: 7.4 baseSeverity: HIGH version: 3.1
cweId: CWE-122 (CWE-122 Heap-based Buffer Overflow)
affected: Ruby (ruby-lang), defaultStatus unaffected
version 0, lessThan 3.3.14
version 3.4.0, lessThan 3.4.13
version 4.0.0, lessThan 4.0.10
SSVC (CISA-ADP, 2026-11-19T18:55:33Z): Exploitation none, Automatable no, Technical Impact total
The second block is Veria's affected and fixed table as printed, with the pipe characters added by this site to show the columns.
Ruby series | Affected | Fixed in
4.0 | 4.0.0 to 4.0.9 | 4.0.10
3.4 | 3.4.0 to 3.4.12 | 3.4.13
3.3 | 3.3.0 to 3.3.13 | 3.3.14
3.2 and older | All releases back to 2.4.0 | EOL
The third block is the Ruby advisory's entry for the flaw the record links, with the affected versions and the recommended action. The same ranges apply to all seven entries, and the advisory prints no CVE id.
GHSA-34m4-m9fh-gqp5: Heap buffer overflow in case conversion of strings
Affected versions
Ruby 3.3.13 and earlier
Ruby 3.4.0 through 3.4.12
Ruby 4.0.0 through 4.0.9
Recommended action
We strongly recommend upgrading to Ruby 4.0.10, 3.4.13 or 3.3.14.
Table 7. Veria's framework status table as printed ("Status on an affected Ruby").
| Framework | Status on an affected Ruby | Veria's qualifier |
|---|---|---|
| Rails | Affected | In their default configuration |
| Sinatra (classic style) | Affected | In their default configuration |
| Other frameworks and plain Rack | Not exposed by default | Application code that upcases strings received in a legacy encoding may also be vulnerable |
The operations that Veria and Previdian both say reach the flaw are listed in the next block. They are the names to search for when reviewing application code, as watchTowr suggests ("Review application code for functions that change the case of text coming from outside the application"). Both pages say UTF-8 strings take a different code path and are not affected by this flaw. The list gives a reviewer a starting point and not a test: the pages do not say which calls on which inputs reach the flaw, and a search for the names will also match harmless uses.
upcase, swapcase, downcase(:fold) and casecmp?
Table 8. Fixed versions the saved pages print, for the framework and bundled software cases.
| Where | Fixed version printed | Printed by |
|---|---|---|
| Ruby 4.0 series | 4.0.10 | Ruby, record, Veria |
| Ruby 3.4 series | 3.4.13 | Ruby, record, Veria |
| Ruby 3.3 series | 3.3.14 | Ruby, record, Veria |
| Discourse base image | 2.0.20261119-0203 | Veria (Status: Patched) |
| GitHub Enterprise Server | None | Veria (Status: Vulnerable) |
What to do, in order
The first line is what the maintainers and Veria say: upgrade to Ruby 4.0.10, 3.4.13 or 3.3.14. The rest is this site's suggestion, and the saved pages do not say how any given estate is built.
Take this with you
A checklist for anyone who runs Ruby
- Upgrade to Ruby 4.0.10, 3.4.13 or 3.3.14 and restart (Ruby, Veria). Rebuild container images and check the Ruby that is actually running, not the one installed beside it (Previdian, watchTowr).
- Find every Ruby runtime, including the ones inside vendor products and container base images, background workers and bundled runtimes (Veria says the fix has to come from the vendor for products that bundle their own Ruby). This site's suggestion: write the list down before the proof of concept date.
- Prioritise Rails and classic style Sinatra applications, the two named default exposures (Veria), and any code that changes the case of text from outside the application (watchTowr). This site's suggestion: that is the review to start now, with the four operation names above.
- Ruby 3.2 and older gets no fix (Ruby). Plan the move to a supported series; until then, the only mitigation Previdian prints is restricting public access to the affected services, and it says no vendor confirmed workaround has been published.
- Self hosted products that bundle Ruby: check the vendor's own release notes and status, not only Veria's table. GitHub Enterprise Server was marked Vulnerable with no fixed release in Veria's table on 20 November.
- If a scanner or ticket matches on CVE ids, re-run it now that one exists, and add the GHSA ids for the other six. This site's suggestion: key the inventory on version, not on the id, until Ruby's page and the record agree.
- Do not treat the 7.4 score as the whole picture, and do not treat the word Critical as one either. This site's suggestion: rank the work by whether the application is internet facing and changes the case of outside text, which is what the pages actually describe.
- Watch for Veria's public proof of concept, planned for 25 November or later, and expect the risk to move when it appears. Nothing saved says what changes then.
The question the maintainers' page leaves open
The CVE record exists, links the GHSA entry, and scores the flaw 7.4 High with a local vector. The maintainers' page, a day and more later, still says no id is assigned and prints no score. Veria says Critical, with network access and no authentication, and promises a public proof of concept in five days. If the maintainers' page lags the record, and the record's vector disagrees with the finder's header, which page does your scanner read, and which severity does your queue sort on?
Key facts
Sources
- PrimarySecurity advisories: multiple vulnerabilities in Regexp and String case conversion, dated 18 November 2026, History line 19 November 00:00 UTC. Re-read in full at about 15:19 UTC on 20 November 2026; text identical to the 19 November copyRuby (ruby-lang.org)accessed 2026-11-20
- PrimaryCVE-2026-132825: description, CVSS 3.1 vector and score 7.4, affected ranges, references, CISA-ADP SSVC rating. Read in full on 20 November 2026; record last updated 20 November 04:06 UTCCVE Services (record assigned by mitre, with CISA-ADP and CVE Program containers)accessed 2026-11-20
- PrimaryRuby2Shell: affected and fixed versions, framework status table, self hosted software table, header (Severity Critical, Network, no authentication), PoC date. Page says Updated 20 November 2026 at 15:00 UTC; saved at about 15:18 UTC. The technical section was read and is not reproducedVeria Labsaccessed 2026-11-20
- PrimaryRuby2Shell (CVE-2026-132825): what we know so far, published and last updated 19 November 2026: severity line, exploitation status, product list as of 19 November. A vendor page, used for its own claimsPrevidianaccessed 2026-11-20
- PrimaryRuby2Shell (CVE-2026-132825) FAQ, published 19 November 2026 at 20:54 UTC: conditions, version table, fix guidance. A vendor page, used for its own claimswatchTowraccessed 2026-11-20
- Reported byRuby Patches Ruby2Shell Vulnerability That Could Lead to RCE, 20 November 2026: relays a watchTowr specialist on exploitation difficulty and lists product statuses. Secondary; the quoted specialist and the reporter are not named hereSecurityWeekaccessed 2026-11-20


