BlackTreeCVE Intelligence
← Back to the CVE catalogue
Full vulnerability report · 2026
CVE-2026-43632High confidence

llama.cpp b7492–b9060 Use-After-Free in Tokenization Endpoints

ggml-org · llama.cpp

9.2CriticalCVSS 4.0
Recommended action
Within 72 hours

Critical technical impact with a remotely reachable, unauthenticated path; no CISA KEV confirmation is currently recorded.

Patch available
R
Operational reassessment

Published severity in operational context

Open reassessment dashboard →
Published severityCriticalOperational priority:Critical, unchanged from published severity.unchanged

Evidence used

  • No CISA KEV confirmation is currently recorded.
  • The selected CVSS metric records a network-reachable, unauthenticated path with no user interaction.
  • EPSS is 0.48% for the current model date.

Compensating controls

  • Validate the affected product branch and deploy the verified fixed release.
  • Restrict the affected network interface to trusted sources where business-safe.
  • Monitor vendor guidance and exploitation sources for a material change.

Verification

  1. Confirm that the asset runs ggml-org llama.cpp and falls inside the recorded affected range.
  2. Verify the installed build against the product-specific fixed version after deployment.
  3. Validate exposure, authentication requirements and compensating controls in the actual environment.
  4. Reopen this reassessment when CVSS, KEV, EPSS, exploit evidence or remediation changes.
Mitigation target: Within 3 daysRemediation target: Within 90 days

This automated reassessment organises public evidence. It does not know asset exposure, business impact or control effectiveness and does not replace CVSS or a human risk decision.

Distribution package intelligence

Release-specific package status

Debian findings are scoped to the named distribution, release and source package. An absent finding does not mean a package is unaffected.

2 package states
Repository candidate not checked

A published vendor fix does not prove that a matching update is enabled and installable on a particular asset. Confirm the local package candidate before scheduling remediation.

Distribution releaseSource packageVendor stateFixed versionEvidence
Debian forkyforky · sourcellama.cppUnder evaluationDebian has not confirmed a fixed package state for this release.Not published in this feedDebian Security Tracker ↗Source updated 5 Oct 2026
Debian sidsid · sourcellama.cppUnder evaluationDebian has not confirmed a fixed package state for this release.Not published in this feedDebian Security Tracker ↗Source updated 5 Oct 2026
Optional official sources

National CERT insights
?CERT means Computer Emergency Response Team; CSIRT is the closely related term Computer Security Incident Response Team.

Choose official national sources for this report. Each advisory shows its original language. Your selection is remembered on this device and included in shared links.

Official European source

ENISA European Vulnerability Database

Official EUVD identifiers, advisory evidence and known-exploited context. Missing fields are not treated as evidence of low risk.

1 current
ENISA EUVD identifier

EUVD-2026-54287

No EUVD known-exploited evidence

ENISA has published the identifier mapping but no EUVD description has been stored yet.

EUVD state
Present in the current official mapping
Known exploitation
Not present in the current ENISA EUVD known-exploited dataset. This is not proof of no exploitation.
ENISA score
Not supplied in the stored EUVD record
Advisory evidence
No linked advisory details stored yet
Recommended actionWithin 72 hours

Critical technical impact with a remotely reachable, unauthenticated path; no CISA KEV confirmation is currently recorded.

Patch available
01

What, why and how

llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in llama-server affecting six tokenization endpoints (/tokenize, /detokenize, /infill, /apply-template, /rerank, and /anthropic/count_tokens) that bypass the task queue and access ctx_server.vocab directly on HTTP worker threads. Attackers can exploit a time-of-check-time-of-use race condition where the main thread destroys and frees vocab after the synchronization lock is released but before the handler finishes using it, causing a crash or potential code execution when --sleep-idle-seconds is configured.

What

llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in llama-server affecting six tokenization endpoints (/tokenize, /detokenize, /infill, /apply-template, /rerank, and /anthropic/count_tokens) that bypass the task queue and access ctx_server.vocab directly on HTTP worker threads. Attackers can exploit a time-of-check-time-of-use race condition where the main thread destroys and frees vocab after the synchronization lock is released but before the handler finishes using it, causing a crash or potential code execution when --sleep-idle-seconds is configured.

Why

The program can continue using memory after it has been released, producing unsafe and attacker-influenceable behaviour.

How

An attacker operating through a network path may attempt exploitation without authentication or user interaction. If successful, the issue may disrupt the affected service.

What

llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in llama-server affecting six tokenization endpoints (/tokenize, /detokenize, /infill, /apply-template, /rerank, and /anthropic/count_tokens) that bypass the task queue and access ctx_server.vocab directly on HTTP worker threads. Attackers can exploit a time-of-check-time-of-use race condition where the main thread destroys and frees vocab after the synchronization lock is released but before the handler finishes using it, causing a crash or potential code execution when --sleep-idle-seconds is configured.

Why

The program can continue using memory after it has been released, producing unsafe and attacker-influenceable behaviour.

How

An attacker operating through a network path may attempt exploitation without authentication or user interaction. If successful, the issue may disrupt the affected service.

02

Exploit reality and attack path

CVSS severity, EPSS forecast probability, public exploit material and CISA-confirmed exploitation are separate signals.

Observed exploitation
?Confirmed exploitation and public exploit material are separate signals. Attacks can occur without public proof-of-concept or exploit code.
No confirmed evidence

No CISA KEV match was present at the last successful refresh. This means no confirmation from that source, not proof of no exploitation.

Public PoC / exploit material
?Confirmed exploitation and public exploit material are separate signals. Attacks can occur without public proof-of-concept or exploit code.
None recorded

No exploit-tagged reference or CISA SSVC proof-of-concept state is currently recorded. Research may still exist outside the structured feeds.

Likely attack path
a network path → Use After Free → disrupt the affected service
Attack surface
Network
Privileges required
None: unauthenticated exploitation is possible
User interaction
None
Attack complexity
High: exploitation depends on specific conditions
Security boundary
Not a CVSS 4.0 base metric
Weakness
?CWE means Common Weakness Enumeration: a standard category for the underlying weakness.
CWE-416 ↗

CWE-416: Use After Free. The product reuses or references memory after it has been freed. At some point afterward, the memory may be allocated again and saved in another pointer, while the original pointer references a location somewhere within the new allocation. Any operations using the original pointer are no longer valid because the memory belongs to the code that operates on the new pointer.

CVSS vector
?CVSS means Common Vulnerability Scoring System. The vector records the metric values used to calculate technical severity.
CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N

Common Vulnerability Scoring System 4.0: the compact vector below is decoded into plain language.

AVNetworkAttack vector: The vulnerable component can be reached over a network.ACHighAttack complexity: Successful exploitation depends on specific conditions outside the attacker's direct control.ATPresentAttack requirements: A particular deployment or execution condition must be present.PRNonePrivileges required: The attacker does not need an account or existing privileges.UINoneUser interaction: No action by another user is required.VCHighVulnerable-system confidentiality: A successful attack can cause a major loss.VIHighVulnerable-system integrity: A successful attack can cause a major loss.VAHighVulnerable-system availability: A successful attack can cause a major loss.SCNoneSubsequent-system confidentiality: No direct loss is represented by this metric.SINoneSubsequent-system integrity: No direct loss is represented by this metric.SANoneSubsequent-system availability: No direct loss is represented by this metric.
Post-exploitation / living off the land
Stolen credentials, tokens and legitimate administration functions may provide continued access without deploying a large custom toolset.
NetworkUnauthenticatedDenial of serviceCWE-416
A

Official authority intelligence

Only matched European and national findings are included. Language selectors and unavailable sources are omitted.

Cyber Security Agency of Singapore · English · CSA-SB-20260812Security Bulletin 12 Aug 2026

The Cyber Security Agency of Singapore included this CVE in its official Security Bulletin 12 Aug 2026, published on 12 August 2026. Open the linked bulletin for the product, severity and reference information published in that issue.

Official advisory ↗
JVN iPedia · Japanese · JVNDB-2026-035217ggml.aiのllama.cppにおける複数の脆弱性

llama.cppのb1283からb9058のビルドには、llama_batch_init()関数に整数オーバーフローの脆弱性が含まれています。この関数内でのmalloc()呼び出し時に、チェックされていない乗算が行われるため、割り当てサイズの計算時にINT32_MAXを超えてオーバーフローが発生する可能性があります。攻撃者は特別に細工されたパラメータを渡すことで整数オーバーフローを引き起こし、ヒープの破損を誘発します。その後のバッチ操作で割り当てられたバッファ境界を越えて書き込みを行うことで、任意のコード実行を達成する可能性があります。

Official advisory ↗
JVN iPedia · Japanese · JVNDB-2026-035216ggml.aiのllama.cppにおける複数の脆弱性

llama.cppのビルドb3978からb9058には、DRYサンプラーに整数のアンダーフローおよび境界外読み取りの脆弱性が存在します。この脆弱性により、認証されていない攻撃者がdry_allowed_lengthをINT32_MINに設定した細工されたHTTPリクエストを/v1/completionsまたは/v1/chat/completionsエンドポイントに送信することで、ヒープバッファのアンダーフローを引き起こすことが可能です。攻撃者はこの脆弱性を悪用して、SIGSEGVによるサーバークラッシュを引き起こし、接続中のすべてのユーザーに対するサービス拒否(DoS)を発生させたり、割り当てられたバッファ前のメモリからゴミ値を読み取ってトークンサンプリング確率を破損させたりできます。

Official advisory ↗
JVN iPedia · Japanese · JVNDB-2026-035215ggml.aiのllama.cppにおける複数の脆弱性

llama.cppのb4882からb9058までのビルドには、KVキャッシュ状態の復元経路にヒープバッファオーバーフローの脆弱性があります。これはstate_read_data()関数がオーバーフローチェックを行わずに書き込みサイズを計算するためであり、slot_save_pathディレクトリへの書き込み権限を持つ攻撃者がヒープメモリを破損させる可能性があります。攻撃者はcell_countの乗算でオーバーフローを発生させるか、テンソルバッファの割り当てを超える悪意のある状態ファイルを作成することで、バッファ境界を越えて攻撃者が制御するバイトを書き込むことができ、その結果ヒープメタデータが破損したり、モデル重みが破損したり、関数ポインタが上書きされて任意コード実行が引き起こされる可能性があります。

Official advisory ↗
JVN iPedia · Japanese · JVNDB-2026-035213ggml.aiのllama.cppにおける複数の脆弱性

llama.cppのビルドb7492から最新のb9060までの間で、--sleep-idle-seconds機能が有効化されている場合に、llama-serverのvocabポインタにUse-After-Freeの脆弱性が存在します。これにより、認証されていないリモート攻撃者が任意のコードを実行できる可能性があります。攻撃者は、サーバーがスリープモードに移行する際に影響を受けるエンドポイントにリクエストを送信することで脆弱性を引き起こせます。その結果、同時に実行されているワーカースレッドが解放されたvocabポインタを逆参照し、そのポインタが攻撃者が制御可能なデータで再利用されることで、リモートコード実行が達成されます。

Official advisory ↗
JVN iPedia · Japanese · JVNDB-2026-035212ggml.aiのllama.cppにおける複数の脆弱性

llama.cppのビルドb7492から最新のb9060にかけて、llama-serverにuse-after-freeの脆弱性が含まれています。この脆弱性は、6つのトークン化エンドポイント(/tokenize、/detokenize、/infill、/apply-template、/rerank、および/anthropic/count_tokens)に影響を及ぼします。これらのエンドポイントはタスクキューをバイパスしてHTTPワーカースレッド上で直接ctx_server.vocabにアクセスします。攻撃者は、メインスレッドが同期ロックを解除した後にvocabを破棄・解放し、その後ハンドラが使用を完了する前に、チェック時と使用時の競合状態(time-of-check-time-of-use)を悪用することが可能です。これにより、--sleep-idle-secondsが設定されている場合にクラッシュや潜在的なコード実行が発生する可能性があります。

Official advisory ↗
JVN iPedia · Japanese · JVNDB-2026-035214ggml.aiのllama.cppにおける複数の脆弱性

llama.cpp の b5702 から b7653 のビルドには、再帰的メモリ状態復元パスにおける境界外読み取りの脆弱性が含まれています。これにより、スロット保存ディレクトリへの書き込み権限を持つ攻撃者が、割り当てられたセル配列の末尾を超えるメモリを読み取ることが可能になります。攻撃者は、過剰なサイズの seq_id 値を持つ悪意のあるスロットファイルを作成することで境界外読み取りを引き起こし、ポインタ値を含むヒープデータをサーバーログに漏洩させます。これにより ASLR 保護が無効化され、更なる攻撃を容易にします。

Official advisory ↗
03

Patch and workaround

Operational remediation based on structured source evidence.

Status
?Patch availability is based on structured fixed-version fields and authoritative update references. If no fix is verified, check the vendor advisory before making a change.
Patch available
Affected
llama.cpp: b7492 ≤ b9060
Fixed
An authoritative update reference is available, but the fixed version is not recorded in the structured CVE fields. Check the linked vendor advisory for the applicable release.
Action
Review the linked authoritative reference and apply the recorded fixed release appropriate to the affected product branch.
Workaround
No verified workaround is recorded. If business-safe, reduce exposure to the affected interface and allow only trusted sources until authoritative guidance is available.
04

Evidence and provenance

Published 6 Aug 2026 · Last source change 14 Aug 2026, 16:49 UTC · CWE-416 · Use After Free

CVE recordCVE.org · 5.2
CVSS sourceCNA
EPSS source
?The date BlackTree first stored a score for this CVE from the daily FIRST EPSS feed.
FIRST · tracked since 2026-08-14
European sourceENISA EUVD · EUVD-2026-54287
Product sourceCNA
Remediation sourceCVE/CNA references
CWE sourceCNA
NVD statusNVD enriched

Core structured fields are present and their contributing authorities are shown above.

Material change intelligence

What changed after publication

View recent updates ↗
  1. Vendor guidanceAuthoritative vendor guidance changed: added patch: github.com/llama-cpp-security-patches; added patch: vulncheck.com/llama-cpp-b7492-b9060-use-after-free-in-tokenization-endpoints; 1 further additions.
    Before
    no authoritative guidance recorded
    After
    patch: github.com/llama-cpp-security-patches · patch: vulncheck.com/llama-cpp-b7492-b9060-use-after-free-in-tokenization-endpoints · remediation: github.com/llama-cpp-security-patches
    disclosure@vulncheck.com ↗
  2. Affected versionsThe structured affected or fixed version information changed.
    Before
    b7492 ≤ b9060 · Fixed: No fixed version is explicitly recorded in the structured CVE data.
    After
    llama.cpp: b7492 ≤ b9060 · Fixed: An authoritative update reference is available, but the fixed version is not recorded in the structured CVE fields. Check the linked vendor advisory for the applicable release.
    CNA ↗
  3. Remediation statusRemediation status changed from Awaiting fix to Patch available.
    Before
    Awaiting fix
    After
    Patch available
    disclosure@vulncheck.com ↗
Material fields only · duplicate refreshes suppressed · history retained for the configured operational retention period
Technical terms and abbreviations used in this report
CVE
Common Vulnerabilities and Exposures: the public identifier for one disclosed vulnerability.
CVSS
Common Vulnerability Scoring System: a technical severity framework; it is not patching priority by itself.
EPSS
Exploit Prediction Scoring System: FIRST's estimate of the probability that exploitation activity will be observed in the next 30 days; it is a forecast, not confirmation.
CWE
Common Weakness Enumeration: the standard category describing the underlying software or hardware weakness.
CNA
CVE Numbering Authority: an organisation authorised to assign and publish CVE records.
CISA ADP
Cybersecurity and Infrastructure Security Agency Authorized Data Publisher: structured enrichment added to a CVE record.
NVD
National Vulnerability Database: NIST's enrichment service for CVE records.
CERT / CSIRT
A computer security incident response team that publishes warnings or coordinates incident response.
PoC
Proof of concept: public material that demonstrates or helps reproduce exploitation.
CSAF
Common Security Advisory Framework: a machine-readable format for security advisories.
LoTL
Living off the land: abuse of legitimate tools or system functions during an attack.
Free version - for non-commercial use only.CVE-2026-43632 · cve.blacktree.nl