Any plans to address known vulnerability identified in CVE-2026-13500 / BDSA-2026-17031?

[RainerPodlas]

Dear All,

We are using recent ANTLR4 4.13.2. Some weeks ago we have been informed (by a Blackduck vulnerability scan) about a severe weakness we need to address:

A weakness has been identified in antlr ANTLR4 up to 4.13.2. Affected is an unknown function of the file tool/src/org/antlr/v4/codegen/model/OutputFile.java of the component Grammar Action Block Handler. Executing a manipulation can lead to code injection. The attack may be launched remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.

Are there any plans to resolve this vulnerability? If so, when will there be an update available?
Feedback is very much appreciated.

Thank you very much.
All the best, Rainer

[ericvergnaud]

Hi there,Can you point us to the specific line that is causing that alert?Kind regardsEnvoyé de mon iPhoneLe 20 juil. 2026 à 09:33, Rainer Podlas ***@***.***> a écrit : Dear All, We are using recent ANTLR4 4.13.2. Some weeks ago we have been informed (by a Blackduck vulnerability scan) about a severe weakness we need to address: A weakness has been identified in antlr ANTLR4 up to 4.13.2. Affected is an unknown function of the file tool/src/org/antlr/v4/codegen/model/OutputFile.java of the component Grammar Action Block Handler. Executing a manipulation can lead to code injection. The attack may be launched remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way. Are there any plans to resolve this vulnerability? If so, when will there be an update available? Feedback is very much appreciated. Thank you very much. All the best, Rainer —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[RainerPodlas]

Dear Eric,
I could not directly read it from the Blackduck scan result, but I did some research and found the following sources describing the issue in more detail:

• antlr4_001_Code_Injection_via_Grammar_Actions wooyun123/wooyun#4
• https://vuldb.com/cve/CVE-2026-13500
• https://vuldb.com/submit/838568
• GHSA-m64j-m5c8-f58p

It seems lines 39–47 are the problematic ones. The implementation in "buildNamedActions(Grammar g, Predicate filter)".
Hope this helps.

Please let me know.
Thank you!

All the best,
Rainer

[ericvergnaud]

Hi,

thanks for this, the provided details are useful.

As I understand it, here is the reasoning:

Context:

• a g4 file may contain user code (predicates and/or action code)
• the tool integrates that user code into the generated lexer or parser

Risk:

• an attacker could inject malicious code in a g4
• the tool would the generate lexer or parser containing malicious code
• when executed, the lexer and parser could execute malicious code

If the above correctly reflects the finding, then I'm not sure it's possible to address this security issue without breaking the product...
I would tend to think that for such feature, preventing code injection would be the responsibility of the people hosting the tool?

@parrt @kaby76 what do you think?

[kaby76]

I'm not sure even where to begin or how to do the scrubbing of action blocks. Generate the code then look at the generated code?? Punt.

[ericvergnaud]

I guess the only way would be to create our own language for predicates and simple actions. I don't see how we could possibly scrub code that uses imports we don't have access to...
(@parrt I'm NOT proposing that we do that!!!)

[parrt]

Yeah I don't see how we could scrub those. Sorry!

…

On Mon, Jul 20, 2026 at 5:59 AM Eric Vergnaud ***@***.***> wrote: I guess the only way would be to create our own language for predicates and simple actions. I don't see how we could possibly scrub code that uses imports we don't have access to... ***@***.*** <https://github.com/parrt> I'm NOT proposing that we do that!!!) — Reply to this email directly, view it on GitHub <#4956?email_source=notifications&email_token=AABLUWMIVT434UDDDVTDKA35FYJSZA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZXGAYTONJUUZZGKYLTN5XKO3LFNZ2GS33OUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#discussioncomment-17701754>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABLUWOAF4XAZ5XO3KW6RG35FYJSZAVCNFSNUABFKJSXA33TNF2G64TZHM2TAMJWHA3TWRDJONRXK43TNFXW4OZRGA2DMMBRGI42C5QC> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/AABLUWO6LHJKWCBGLSVTF6D5FYJSZA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZXGAYTONJUUZZGKYLTN5XKO3LFNZ2GS33OUVSXMZLOOSVGM33PORSXEX3JN5ZQ> and Android <https://github.com/notifications/mobile/android/AABLUWJKFXGU5GMQH5YVPG35FYJSZA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZXGAYTONJUUZZGKYLTN5XKO3LFNZ2GS33OUVSXMZLOOSXGM33PORSXEX3BNZSHE33JMQ>. Download it today! You are receiving this because you were mentioned.Message ID: ***@***.***>

-- Dictation in use. Please excuse homophones, malapropisms, and nonsense.

[RainerPodlas]

Thank you all for your answers!

So this threat is only real when using predicates or simple actions. I see that you cannot guarantee that no malicious code is injected by them. How about the possibility to "disable" them on demand (to run ANTLR without these features), so that users who do not need/use them are safe and can argue that? I am not so familiar with ANTLR, but what I have read online, users tend to advise against using action blocks and recommend using visitors and listeners instead...

Have you actually been contacted by the reporter of the CVE record and have you shared your thoughts about it? (The ticket says: "The vendor was contacted early about this disclosure but did not respond in any way.")

Maybe your thoughts and feedback can also help to downgrade the CVE? Just an idea... :-)

All the best,
Rainer

[ericvergnaud]

I guess this might be it: #4952 ?