
Offline scanner telling you which of the 2026 Bouncy Castle CVEs actually apply to you - across BC, BC-LTS and BC-FJA (FIPS), which do not share a version scheme. CVE-2026-58062 / CVE-2026-8763 / CVE-2026-59650 / CVE-2026-59638
Offline triage of which of the 2026 batch of Bouncy Castle CVEs affect your Java application. Zero dependencies, single jar, no network access, no data uploaded.
java -jar bc-check.jar ./app.jar
java -jar bc-check.jar --gav bcprov-lts8on:2.73.5 --detail
All official CVE descriptions uniformly say one sentence:
Bouncy Castle for Java before 1.85
But among the people actually using it, two product lines have version numbers that don't look like this at all:
| Product line | artifact looks like | version number looks like | Is "before 1.85" useful for it |
|---|---|---|---|
| BC (regular edition) | bcprov-jdk18on | 1.81 | ✅ Useful |
| BC-LTS (long-term support edition) | bcprov-lts8on | 2.73.5 | ❌ Completely impossible to judge |
| BC-FJA (FIPS certified edition) | bc-fips / bctls-fips … | 1.0.2.5 / 2.0.1 / 2.1.2 | ❌ Completely impossible to judge |
The fixed versions for the three lines are 1.85, 2.73.12, and a separate set of version numbers for each of the 7 FIPS artifacts. They are not on the same number line, and comparing them against each other is meaningless.
CVE-2026-8149 does not affect the BC line at all, and only affects BC-LTS 2.73.0–2.73.10.
People staring at the version number "1.85" will never find it.CVE-2026-59650 does not affect the FIPS version at all. Reporting it based on "before 1.85" is a false alarm.CVE-2026-59638, the artifact to check on the FIPS side is bctls-fips, not bc-fips ——
it moved from the crypto core to the TLS module. People only watching bc-fips won't find it.CVE-2026-58062 and 58060, the official listing only includes 2.0.2 and 2.1.3 of bc-fips, with no 1.0.x ——
meaning the 1.0.x line is not affected. But CVE-2026-8763 lists 1.0.2.7, so 1.0.x is affected.
Same artifact, same version, different CVEs give different answers.There are also 10 CVEs with a starting affected version (from 1.61, from 1.66, from 1.73 all differ),
so it is not "all versions affected"; another 4 older CVEs have fixed versions that are branch versions like 1.80.2 / 1.81.1 / 1.84,
and the affected ranges are disjoint (1.74–1.80.1, 1.81, 1.82–1.83).
Suppose you have already read all the advisories and know all the answers — you still can't check —— because the answer is not a single version number, but a three-dimensional table of "product line × artifact × CVE".
Correction (2026-10-12): This section originally said "all 4 CVSS critical CVEs are
unreviewed, and Dependabot can't see them". Among them,CVE-2026-8763was reviewed and included by GitHub on 2026-09-18, with package information added, so Dependabot and OSV can now see it. What remains invisible are the other 3. Below is the version revised based on actual testing on 2026-10-12.
Of the 4 CVSS critical CVEs, 3 still have the status unreviewed in the GitHub Advisory,
and their affected package lists are empty (package count 0). Without package names, there is no way to match against the dependency tree ——
they don't enter OSV, Dependabot doesn't alert, and SCA scanning of dependency manifests can't find them either.
Querying OSV with org.bouncycastle:bcprov-jdk18on:1.81 (2026-10-12) returns 4 hits:
CVE-2025-14813, CVE-2026-0636, CVE-2026-13506, CVE-2026-8763 ——
none of the other 3 critical ones are included.
⚠️ To be clear about the boundary: this is not "Dependabot can't see any BC vulnerabilities" ——
CVE-2026-0636in the same batch is reviewed, has package information, and alerts normally. What is invisible are the 3 critical ones markedunreviewedin the table below.
| CVE | GHSA | Status (2026-10-12) | Package count |
|---|---|---|---|
| CVE-2026-58062 | GHSA-j295-77c3-9frf | unreviewed | 0 |
| CVE-2026-8763 | GHSA-9pwp-9qqc-pr26 | reviewed (since 2026-09-18) | 6 |
| CVE-2026-59650 | GHSA-ghgq-7g74-28wp | unreviewed | 0 |
| CVE-2026-59638 | GHSA-8rxj-3p7p-rq39 | unreviewed | 0 |
The fixed versions registered in the global database for CVE-2026-8763 are bcprov-jdk18on / bcprov-jdk15to18 1.85,
bcprov-lts8on 2.73.12, and bc-fips 1.0.2.7 / 2.0.2 / 2.1.3, consistent with this tool's determination table.
java -jar bc-check.jar <路径> [更多路径...] [选项]
java -jar bc-check.jar --gav <坐标>
The path can be a jar / war file (including Spring Boot fat-JARs, automatically expanded layer by layer), or a directory (recursively scanned).
| Option | Description |
|---|---|
--gav <coordinates> | Do not scan files; directly evaluate a single coordinate, e.g. bcprov-lts8on:2.73.5 |
-d, --detail | List each matched CVE and the official original text |
--json | Output JSON for easy pipeline integration (always UTF-8) |
--utf8 / --gbk | Force the specified encoding when console Chinese text is garbled |
--no-color | Disable colored output |
Exit codes: 0 = no match; 1 = CVE matched; 2 = usage error. Can be used directly as a CI gate.
[CRITICAL] bcprov-lts8on 2.73.5 BC-LTS 线
位置 :app.jar!/BOOT-INF/lib/bcprov-lts8on-2.73.5.jar
来源 :jar 内 MANIFEST
结论 :命中 29 条 CVE,其中 CVSS critical 4 条
降噪 :其中 13 条按官方 FIPS 侧点名的 artifact 反推,大概率出在 bcpkix / bcpg / bctls 模块。
⚠️ 这是推断不是官方结论,仅供排优先级。
处置 :升级到 2.73.12(BC-LTS 线)
In the vast majority of cases, BC is not something you pulled in yourself — it gets dragged in as a transitive dependency. TLS, certificate validation, JWT, PDF signing, and PGP all pull it in. So this tool doesn't look at the pom; it only looks at the real artifacts:
Bundle-SymbolicName + Bundle-Version in META-INF/MANIFEST.MF (primary path)META-INF/maven/org.bouncycastle/*/pom.properties (not present in official jars, only in repackaged ones)org/bouncycastle/ classes found but coordinates can't be identified → reports UNKNOWN and prompts for manual confirmationCase 4 is when it has been shaded into the host jar; mvn dependency:tree and SCA can't see it,
this tool will at least tell you "there's BC here, but I can't identify the version," rather than silently missing it.
When a jar has been renamed (common with internal artifact repositories re-releasing), the file name can't be identified, but the MANIFEST still can.
All of it comes from the per-CVE subpages for the 37 CVEs in 2026 on the official github.com/bcgit/bc-java wiki
(the original text of Issue affecting / Fixed versions), not secondhand retellings, not aggregated articles.
tools/data/bc_cves.jsonCveTable.java by tools/gen_rules.py, not copied by handRegenerate: python tools/gen_rules.py
This tool only covers the 37 CVEs in 2026 on the official wiki. Bouncy Castle vulnerabilities before 2026 are out of scope — most of those are already in OSV, findable by Dependabot / SCA, and this tool doesn't reinvent the wheel.