Ruby says 7 Regexp and String flaws are fixed in 4.0.10, 3.4.13 and 3.3.14; PoCs known for 6, no CVE ids yet
Ruby's maintainers say 7 flaws in the regular expression engine and string case conversion are fixed in 4.0.10, 3.4.13 and 3.3.14, with PoC statements for 6 and no CVE ids yet. Ruby 3.2 and earlier get no fix, and the advisory prints no condition for its two case conversion flaws.
By Parminder Kumar Sharma · LinkedIn · · 15 min read
Seven flaws, six PoC statements, 26 days after the last Regexp fixes (derived)
On Wednesday 18 November 2026 Ruby's maintainers published an advisory titled "Security advisories: multiple vulnerabilities in Regexp and String case conversion" and released Ruby 4.0.10, 3.4.13 and 3.3.14. The advisory says seven vulnerabilities were found in Ruby's regular expression engine and in the case conversion of strings, and that "We are aware of PoCs that prove some of them can be used to trigger remote code execution." Counting its entries, six carry a PoC sentence: five say the PoCs prove remote code execution (RCE) "in some scenarios" and one says they prove memory corruption "in an attacker-controlled way". The seventh has none (derived counts).
The previous fixes to the same engine shipped 26 days earlier, on 23 October (derived: 23 October to 18 November). The new advisory's affected ranges end at Ruby 4.0.9, 3.4.12 and 3.3.13, and those three are the releases Ruby's 23 October post says carried the October fixes. So the October releases are inside the new affected ranges (derived by comparing the two pages).
The advisory is the maintainers' own page, and it is the only primary this site saved for the seven flaws. Here is what the counts do not establish:
- Whether anyone is exploiting any of the seven. No saved page says.
- Where the PoCs are, who holds them, or what the "scenarios" are. The advisory does not say.
- Which Ruby methods and inputs reach the two case conversion flaws. Not stated.
- Whether a default Rails or Rack code path calls them on request data. No saved page says, and this brief does not guess.
- A CVE id or a CVSS score for any of the seven. None is printed.
- How many applications are affected. Not stated.
Table 1. The figures this brief opens on. Counts are made by script over the saved advisory; day counts are calendar days (derived).
| Figure | Value | Where it comes from |
|---|---|---|
| Vulnerabilities in the advisory | 7 | The seven GHSA headings. |
| Entries with a PoC sentence | 6 (5 say RCE, 1 says memory corruption) | Counted in the entries; the seventh has none. |
| Entries that say RCE may be possible | 6 of 7 | Counted in the entries; the seventh says memory disclosure or a crash. |
| Entries that print a condition about attacker-controlled patterns | 4 of 7 | Counted; the two case conversion entries and the seventh print none. |
| Days from the October Regexp fixes to this advisory | 26 | 23 Oct 2026 to 18 Nov 2026 (derived). |
| Days from this advisory to the expected end of Ruby 3.3 support | 133 | 18 Nov 2026 to 31 Mar 2027 (derived). |
| Entries credited to Ada Logics and Anthropic | 4 of 7 | Counted in the Credits list. |
The seven entries, side by side
Diagram 1 sets out all seven entries in the order the advisory prints them. Each is identified by a GitHub Security Advisory (GHSA) id only: the advisory says CVE ids "have been requested" but "have not been assigned yet", so it uses the GHSA ids instead. Three of the seven are string flaws or begin with a crafted string (two in case conversion, one in case-insensitive backreference matching). Four are regular expression flaws that print the same condition, applications that handle attacker-controlled patterns.
The advisory describes mechanisms in one sentence each: a one byte heap overflow and an integer overflow that causes a heap overflow in case conversion, a double free, an integer overflow that causes a heap overflow and an out-of-bounds read in the regular expression engine, a heap overflow in Regexp#to_s, and a second out-of-bounds read. It prints no code, no pattern, no string and no indicator, and this brief reproduces none.
Two of the five RCE statements sit in plain string case conversion, and no condition is printed for them
Five of the seven entries carry the sentence that PoCs prove RCE in some scenarios. Two of those five are case conversion flaws. Of the first, the advisory says: "Changing the case of a crafted string may allow remote code execution." The flaw is described as a one byte heap overflow. The second case conversion entry is an integer overflow that causes a heap overflow, and it prints the same sentence about changing the case of a crafted string. Neither prints a condition.
The four regular expression entries that say RCE may be possible do print one. They say it may be possible in "applications that compile attacker-controlled patterns" (the matching entry uses the word "use" in place of "compile"). That is the advisory's own boundary: an application that never compiles or matches a pattern an attacker supplies is outside it, as the advisory frames those four.
The advisory prints no such boundary for string case conversion. It does not say which methods, which encodings or which inputs reach the flaws, and it does not say whether changing the case of ordinary text from a request is enough. Many applications change the case of user text (this site's observation, not the advisory's), but the advisory does not say whether any such use is exposed, so this brief draws no conclusion.
The seventh entry reads: "Matching a crafted string may disclose memory contents or crash the process." It has no PoC sentence and no stated condition about patterns. Its title names case-insensitive backreference matching, which this site reads (inference, not the advisory's statement) as meaning a pattern that uses one. The advisory does not say whether a developer-written pattern is enough or whether the pattern must come from an attacker.
How the October post differed. Ruby's October post said: "Applications that only use regular expressions written by their own developers are not affected." It also said "Regexp.timeout does not protect against these bugs." and gave a workaround: "If you cannot upgrade, please avoid compiling regular expressions from untrusted input." The November advisory, as saved, prints none of those three sentences. For the seven flaws it prints no workaround, and it says nothing about timeouts or pattern length. This site therefore does not extend the October statements to November, and does not claim they no longer hold.
Table 2. What the advisory states about reach, and what it leaves out. Read from the saved advisory of 18 November 2026.
| Question | Stated | Not stated |
|---|---|---|
| Condition for the four regular expression RCE and memory entries | Applications that compile or use attacker-controlled patterns | Whether an application with only developer-written patterns is safe (October post said so for its bugs) |
| Condition for the two case conversion flaws | A crafted string | Which methods, encodings and inputs; whether ordinary request text qualifies |
| Condition for the seventh entry | A crafted string; memory disclosure or a crash | Whether the pattern must be attacker-controlled; whether remote reach exists |
| PoC sentences | Six entries; five say RCE in some scenarios; one says attacker-controlled memory corruption | Which scenarios; where the PoCs are; who has them |
| Workaround | Upgrade to 4.0.10, 3.4.13 or 3.3.14 | Any mitigation for those who cannot upgrade |
The clock: 26 days since October, 133 days to the end of Ruby 3.3 support (derived)
Diagram 2 puts the three dates on one scale. Ruby's October post (copy saved on 26 October) said Ruby 4.0.9, 3.4.12 and 3.3.13 fixed Regexp compilation bugs that GitLab found could lead to remote code execution on GitLab servers. Twenty six days later the advisory says versions up to those same three are affected by seven flaws. The two sets are not said to be related: the November advisory does not mention GitLab or the October fixes.
Ruby 3.3 and its label. The advisory says: "Ruby 3.3 is in the security maintenance phase." The 3.3.14 release post says: "we have renamed this phase from the security maintenance phase to the legacy maintenance phase" and that the support period is unchanged, with official support "still scheduled to end at the end of March 2027". The branches page, as saved on 26 October, prints the end of life as 2027-03-31 (expected), which gives 133 days from 18 November (derived; it takes the last day of March as the end). The release post gives the reason for the new name: Ruby 3.3 now also receives bug fixes judged to need backporting, not only security fixes. A renamed phase is a label. It does not change the end date, and the advisory still uses the old name.
Ruby 3.2 and earlier. The advisory says: "Ruby 3.2 and earlier have reached their end of life and will not receive these fixes." The branches page, as saved on 26 October, prints Ruby 3.2's end of life as 2026-04-01, which is 231 days before the advisory (derived). Whether Linux distributions or other vendors carry their own fixes for 3.2 or earlier is not stated on any saved page. A vendor may or may not; the saved pages cannot tell a reader which.
Which release fixes which series, and what the advisory leaves out
The advisory lists three affected ranges and three fixed releases. It prints one range per series and applies it to all seven flaws: it does not say that any flaw affects only some of the series. Release posts for 4.0.10 and 3.4.13 each say the release includes security fixes and point to the advisory; each also says it includes other bug fixes, which the posts leave to the GitHub releases. This site did not read the GitHub releases.
The advisory does not name Ruby 4.1. The branches page, as saved on 26 October, lists 4.1 as a preview. Whether the preview is affected is not stated.
The advisory also says nothing about products that bundle their own Ruby. The October post said GitLab "addressed these issues in its own releases by patching the Ruby it ships". Whether any vendor that ships Ruby has issued anything for the seven is not stated on any saved page, and this site did not look.
No CVE ids yet: the advisory keys everything on GHSA ids
The advisory says: "CVE identifiers have been requested for these vulnerabilities but have not been assigned yet" and refers to the flaws by GHSA ids for that reason. The advisory page links each id to a page in Ruby's GitHub repository, as the saved HTML shows. This site did not request those pages, so everything here about the seven comes from Ruby's own advisory, and what the GitHub pages hold is not known.
Ruby's October post took the opposite position. It said: "We have not requested CVE identifiers in Ruby for these fixes, at least for now." and told readers to use GitLab's ids, adding "for example in a security scanner". For the November flaws Ruby asks for its own ids and the ids do not exist yet.
This brief makes no claim about what any scanner, inventory or ticketing tool does with a GHSA id, or with a record that has no CVE id. The only checkable point is that the advisory itself prints no CVE id and no CVSS score, so there is no number from the maintainers to sort a queue by, and no id that a tool keyed on CVE ids could match today. The question for each reader is what their tooling is keyed on.
Who is credited: four of seven name Ada Logics and Anthropic
The Credits list names, for each flaw, who reported it. Four of the seven credit Ada Logics and Anthropic together. Two credit a single reporter handle each, and one credits two named people with no organisation. The advisory names private individuals in that list and this brief prints none of them.
Two organisations have an interest worth stating once each. Anthropic builds and sells AI models, and is credited as a reporter on four of the seven entries. Ada Logics is a security firm that sells its services, and is credited with Anthropic on the same four. The advisory says nothing about how the flaws were found, what either organisation did in the reports, or whether any tool was used. Nothing more is claimed here.
For comparison, brief 736 counted Chrome 158's credit lines on 17 November and found three that credit Anthropic and Ada Logics together, on High severity lines; that is that page's statement, not this one's. A shared credit line in two products' advisories is a fact about the credits, not about any method.
Where the pages differ, unresolved
The saved pages leave five differences. This brief reports them and does not resolve them.
Table 3. Differences between and inside the saved Ruby pages.
| Point | One page prints | Another page prints |
|---|---|---|
| Date of the advisory | The advisory, the three release posts and the index entry say 18 November 2026 | The advisory's footer says Originally published at 2026-11-19 00:00:00 (UTC). The October post's footer held a full time of day, 03:11:22 UTC on 23 October. The reason for a midnight value is not stated. |
| Name of the Ruby 3.3 phase | The advisory: security maintenance phase | The 3.3.14 post: renamed legacy maintenance phase, support still ends at the end of March 2027 |
| Strength of the recommendation | The advisory: strongly recommend upgrading | The October post: recommend upgrading to 4.0.9, 3.4.12 or 3.3.13 or later |
| The October entry's count of CVE ids | The security index lists the 23 October entry under four GitLab CVE ids | The October post says GitLab published four CVEs for the underlying Ruby bugs and that Ruby requested none. |
| What the PoCs prove | Five entries: RCE in some scenarios | One entry: memory corruption in an attacker-controlled way; one entry: no PoC sentence |
What the advisory prints, line by line
The advisory prints no indicators, no log lines, no test input and no pattern. What a defender can copy from it is the list of ids, the affected ranges and the fixed versions. The block below holds those lines as the advisory prints them, so an inventory or a ticket can quote them exactly. The ids are the only keys the advisory gives, since it prints no CVE ids.
GHSA-34m4-m9fh-gqp5: Heap buffer overflow in case conversion of strings
GHSA-7x2q-68qv-6p8x: Double free in regular expression compilation
GHSA-87fm-7w84-j2px: Integer overflow in case conversion of strings
GHSA-wpqw-5qxg-m68r: Out-of-bounds read in case-insensitive regular expression matching
GHSA-3p8q-wmhx-vm2w: Heap buffer overflow in regular expression compilation
GHSA-7whj-cfh7-xgm4: Heap buffer overflow in Regexp#to_s
GHSA-pvgc-g7gj-pv6h: Out-of-bounds read in case-insensitive backreference matching
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.
Two lines are this site's derived additions, not the advisory's: the three releases that fix the flaws are Ruby 4.0.10, 3.4.13 and 3.3.14, all released on 18 November 2026 according to their release posts. The release posts print tarball hashes and sizes; this brief copies none, because the posts are where a reader should take them from.
What to do, in order
The first line is the advisory's own advice. Everything after it is this site's suggestion.
Take this with you
The advisory says: upgrade to Ruby 4.0.10, 3.4.13 or 3.3.14. This site's suggestions follow.
- Find every Ruby runtime you run: servers, containers, CI images, language version managers and the Ruby bundled inside vendor products.
- Upgrade to Ruby 4.0.10, 3.4.13 or 3.3.14 as the advisory says, then restart every process that loaded the old runtime.
- If you run Ruby 3.2 or earlier, plan a move now. The advisory says those versions will not receive these fixes. Ask any distribution that packages your Ruby whether it carries its own fix; the saved pages do not say.
- Until you have upgraded, do not compile or match patterns an attacker can supply, which is the advisory's own condition for the four regular expression flaws, and treat case conversion of untrusted strings as in scope. This site's reading is that the advisory states no condition for the case conversion flaws.
- Watch for CVE ids and CVSS scores, and re-run your scanners when they are assigned. Track the seven by their GHSA ids until then.
- Keep dependency lock files, base image tags and pipeline logs, so you can show which Ruby a given release ran on during this period.
The question the advisory leaves open
Six of seven entries carry a PoC sentence, four have a condition, and none has a CVE id, a score or a date for exploitation. A queue that sorts on identifiers and scores has nothing to sort these by today. If your scanner matches on CVE ids and none exist yet, what is it matching on?
Key facts
Sources
- PrimarySecurity advisories: multiple vulnerabilities in Regexp and String case conversion, posted 18 November 2026: seven GHSA entries, PoC sentences, conditions, affected and fixed versions, credits, History line. Read in full; saved 19 November 2026 at about 12:16 UTCRuby (ruby-lang.org)accessed 2026-11-19
- PrimaryRuby 4.0.10 Released, 18 November 2026: includes security fixes and points to the advisory. Read in full; saved 19 November 2026Ruby (ruby-lang.org)accessed 2026-11-19
- PrimaryRuby 3.4.13 Released, 18 November 2026. Read in full; saved 19 November 2026Ruby (ruby-lang.org)accessed 2026-11-19
- PrimaryRuby 3.3.14 Released, 18 November 2026: security maintenance phase renamed legacy maintenance phase, support still ends at the end of March 2027. Read in full; saved 19 November 2026Ruby (ruby-lang.org)accessed 2026-11-19
- PrimaryRuby security page, list of known issues: the 18 November advisory and the 23 October entry with its four GitLab CVE ids. Saved 19 November 2026Ruby (ruby-lang.org)accessed 2026-11-19
- PrimaryRegexp compilation fixes related to GitLab's CVE-2026-89078, CVE-2026-93577, CVE-2026-97229 and CVE-2026-110264, posted 23 October 2026: Ruby 4.0.9, 3.4.12 and 3.3.13, the condition and workaround text. Read from a copy saved on 26 October 2026 for an earlier audit, not re-read on 19 NovemberRuby (ruby-lang.org)accessed 2026-10-26
- PrimaryRuby Maintenance Branches: Ruby 3.3 expected end of life 2027-03-31, Ruby 3.2 end of life 2026-04-01, Ruby 4.1 preview. Read from a copy saved on 26 October 2026 for an earlier audit, not re-read on 19 NovemberRuby (ruby-lang.org)accessed 2026-10-26


