LiteLLM allows an authenticated internal_user to create API keys with access to routes that their role does not permit

LiteLLM prior to 1.83.14 allows an authenticated internaluser to create API keys with access to routes that their role does not permit. When generating a key, the allowedroutes field is stored without verifying that the specified routes fall within the user's own permissions. A key created with...

Twig: `{% sandbox %}{% include %}` skips checkSecurity() on cached templates (incomplete fix for CVE-2024-45411)

Description The fix for CVE-2024-45411 / GHSA-6j75-5wfj-gh66 added an explicit $loaded-unwrap-checkSecurity call in CoreExtension::include so that a template already cached in Environment::$loadedTemplates is re-checked when included with sandboxed = true. The deprecated but still functional %...

Twig: HTML-output filters in twig/* extras incorrectly declared `is_safe => ['all']`

Description Several filters in the twig/ extras packages are registered with issafe = 'all', which tells Twig's autoescaper to treat their output as safe in every context html, js, css, url, .... The output of these filters is plain text or HTML markup, neither of which is safe in every escaping...

Twig: Sandbox property allowlist bypass via the `column` filter (array_column on objects)

Description The column filter passes its input straight to PHP's native arraycolumn. When the array elements are objects, arraycolumn reads $obj-$name and $obj-$index directly, including invoking get/isset. Because this property read happens entirely in PHP native code and never reaches...

Twig: `template_from_string()` escapes a SourcePolicy-driven sandbox via synthesized template name

Description When the sandbox is enabled selectively via SourcePolicyInterface and not globally, a sandboxed template that is allowed to call templatefromstring and include can render an arbitrary inner template with no security policy enforcement. Environment::createTemplate compiles the inner...

Twig: PHP code injection via `{% use %}` template name

Description Compiler::string escapes ", $, , NUL and TAB when generating PHP double-quoted string literals, but does not escape single quotes. In ModuleNode::compileConstructor, the template name from a % use % tag is compiled via subcompile - string and placed inside a surrounding PHP...

twig/intl-extra: Unbounded formatter memoisation in keyed on template-controlled arguments

Description IntlExtension memoises every \IntlDateFormatter and \NumberFormatter it creates in instance-level arrays keyed on a hash that includes locale, pattern, attrs and other values that are ordinary named arguments of the formatdatetime / formatdate / formattime / formatnumber /...

Twig: The `spaceless` filter implicitly marks its output as safe

Description The spaceless filter is registered with issafe = 'html', which means Twig's autoescaper does not escape its output in an HTML context. As a result, applying spaceless to attacker-controlled input that contains markup emits the markup unescaped even when the developer never wrote |raw...

JavaScript Cookie: Per-instance prototype hijack in assign() enables cookie-attribute injection

Summary js-cookie's internal assign helper copies properties with for...in + plain assignment. When the source object is produced by JSON.parse, the JSON object's "proto" member is an own enumerable property, so the for…in enumerates it and the targetkey = sourcekey write triggers the...

Russh: Unchecked CryptoVec allocation and growth handling is reachable

Title Unchecked CryptoVec allocation and growth handling was reachable from local agent inputs in current russh releases and from remote SSH traffic in historical pre-0.58.0 releases Summary CryptoVec used unchecked capacity growth, unchecked length arithmetic, and unsafe allocation/locking paths...

@hulumi/policies: Stack-wide evidence bypassed Cloudflare and deployment-governance guardrails

Impact: @hulumi/policies versions before 1.3.2 used stack-wide evidence shortcuts in several Cloudflare and deployment-governance validators. Unrelated compliant-looking evidence could suppress violations for different zones, hostnames, origins, or repositories in the same stack. Patched in 1.3.2...

@hulumi/policies: GitHub OIDC trust policy bypass via AWS set-qualified condition operators

Impact: @hulumi/policies versions before 1.3.2 only checked exact AWS IAM StringLike/StringEquals condition operator keys in GOIDC1. Set-qualified operators such as ForAnyValue:StringLike could hide wildcard GitHub Actions OIDC sub conditions from the mandatory guardrail. Patched in 1.3.2: the AW...

@hulumi/policies: CIS 1.16 admin policy bypass for inline and attached IAM policies

Impact: @hulumi/policies versions before 1.3.2 did not fully inspect inline and attached IAM policy evidence for the administrator-policy guardrail, so some admin-equivalent policy paths could pass policy evaluation. Patched in 1.3.2: the validator inspects the affected policy shapes and includes...

@hulumi/policies: HULUMI-H1 SecureBucket parent spoof bypass

Impact: @hulumi/policies versions before 1.3.2 could accept spoofed SecureBucket parent evidence for HULUMI-H1, allowing policy evaluation to miss an unsafe bucket shape. Patched in 1.3.2: the validator now correlates evidence to the expected component/resource relationship and includes regressio...

@hulumi/drift: Orphan reconciler accepted externally supplied execute plans

Impact: @hulumi/drift versions before 1.3.2 could accept externally supplied execute plans without sufficient provenance checks, allowing unsafe reconciliation input to be treated as trusted. Patched in 1.3.2: execute-plan handling now validates provenance and rejects untrusted plans, with...

@hulumi/baseline: CloudTrail selector tampering events were not fully detected

Impact: @hulumi/baseline versions before 1.3.2 could miss some CloudTrail event-selector tampering evidence, reducing coverage for changes to audit logging configuration. Patched in 1.3.2: detection coverage and regression tests were expanded. Remediation: upgrade @hulumi/baseline to 1.3.2 or lat...

Umbraco.Cms: XSS/HTML Injection in Umbraco Backoffice confirmation dialog

Impact Authenticated users are able to inject HTML vulnerability into an input field, which is rendered in the confirmation dialog without proper output encoding. Patches This issue has been patched in 17.4.0...

FlaskBB: SSRF in get_image_info() via unrestricted avatar URL

Summary A Server-Side Request Forgery SSRF vulnerability in getimageinfo allows any authenticated user to force the server to send HTTP requests to arbitrary internal endpoints, including cloud metadata services e.g., AWS 169.254.169.254. This is a blind SSRF with confirmed internal port scanning...

NocoDB: Stale Auth Cache After API Token Deletion

Summary Deleted API tokens continued to authenticate requests until their cache entry expired, because the auth cache was not invalidated by token value at deletion time. Details The API token deletion path removed the database row but did not evict the token-value keyed entry from the auth cache...

NocoDB: Attachment Size Limit Bypass via Upload-by-URL

Summary The upload-by-URL path did not enforce NCATTACHMENTFIELDSIZE against either the remote file's advertised Content-Length or the decoded length of a data: URI, allowing an authenticated user to bypass the configured per-file size limit. Details The attachments service now checks...

NocoDB: Shared-base link access can invite arbitrary users as persistent base members

Summary Shared-base sessions were granted the same base-member capabilities as authenticated viewers. Using only the shared-base UUID xc-shared-base-id, an attacker could enumerate base members and invite an arbitrary email into the base as a real member. The invited user could then redeem the...

NocoDB: Missing File Size Enforcement in Upload-by-URL Allows Denial of Service via Disk Exhaustion

Summary The uploadViaURL path in the v1/v2 attachment API did not enforce NCATTACHMENTFIELDSIZE against the remote content-length or against the response stream. An authenticated user Editor+ could direct the server to download arbitrarily large files, exhausting disk space and causing denial of...

NocoDB: Refresh Token Cookie Set Without `secure` and `sameSite` Flags

Summary The refresh-token cookie was set with httpOnly: true but missing both the secure flag and the sameSite attribute. Over plain HTTP the cookie could be intercepted on the network; without sameSite, browsers attached it to cross-site POSTs, enabling CSRF against the token-refresh endpoint...

NocoDB: OAuth Token Scope Not Enforced at ACL Layer Allows Scope Escalation

Summary The OAuth token strategy attached oauthscope and oauthgrantedresources to the request user, but the ACL middleware never consulted either. An OAuth token issued with a restricted scope e.g. MCP-only therefore inherited the full permissions of the underlying user across all routes; the...

NocoDB: SSRF Protection Bypass in Notification Webhook Plugins (Slack, Discord, Mattermost, Teams)

Summary The request-filtering-agent SSRF protection was non-functional in the four notification webhook plugins Slack, Discord, Mattermost, Teams because httpAgent / httpsAgent were passed as part of the request body rather than the axios config. An authenticated user with hook-creation permissio...

NocoDB: Reflected Cross-Site Scripting via Page Leaving Redirect URL

Summary A reflected XSS vulnerability exists in the Page Leaving Warning page. The ncRedirectUrl and ncBackUrl query parameters are used in window.location.href and tag bindings without validation, allowing javascript: URI injection. Details PageLeavingWarning.vue reads ncRedirectUrl and ncBackUr...

MCP Server Kubernetes: Tool Access Control Bypass via Presentation-Layer Filtering Without Execution-Layer Enforcement

Summary mcp-server-kubernetes exposes three environment variables ALLOWONLYREADONLYTOOLS, ALLOWONLYNONDESTRUCTIVETOOLS, ALLOWEDTOOLS documented as access controls for restricting which Kubernetes operations are available. These controls are enforced at the tool discovery layer tools/list but not ...

SpiceDB: Caveat structures with nested lists can result in improper cache reuse

Impact Users are impacted if: - They have a caveat structure with a nested list, e.g.: zed caveat shapex list x == "a", "b" - Their system exercises that caveat with either CheckBulkPermission or else LookupResources running with the --experimental-lookup-resources-version flag set to lr3, implyi...

Plonky3 MultiField32Challenger: transcript malleability and challenge entropy loss

Impact - Key: challenger/src/multifieldchallenger.rs | MultiField32Challenger::duplexing | transcriptmalleability - Affected files: challenger/src/multifieldchallenger.rs, field/src/helpers.rs - Violated invariant: The Fiat-Shamir sponge must bind challenges to the exact sequence of observed fiel...

Snappy: Binary path is never shell-escaped due to an inverted is_executable check

Impact On POSIX, escapeshellarg‘/usr/bin/wkhtmltopdf’ returns the literal string ‘/usr/bin/wkhtmltopdf’ with the single-quote characters included. isexecutable then looks for a file whose actual name contains those quote characters, which essentially never exists. The safe branch is dead code and...

Snappy : SSRF and local file read via the xsl-style-sheet option

Impact It impacts applications where: - the PHP daemon run with root permissions ; - the application is either running outside a container or has sensitive file access ; It could happens with this kind of workflows: php $stylesheet = $GET'stylesheet'; // = ‘file:///etc/passwd’ $pdf = new...

Fission builder accepts arbitrary buildcmd strings from Environment.spec.builder.command, allowing the builder pod to invoke arbitrary executables

Summary Before the round-1 security sweep, pkg/builder/builder.go passed Environment.spec.builder.command directly into exec.Command... after a strings.Fields split, with no validation of the executable path or its arguments. A user who could create or update Environment CRDs in a namespace...

Fission runtime pods automount the fission-fetcher service-account token into the user function container, granting function code namespace-wide secret / configmap read

Summary Fission runtime pods were created with ServiceAccountName: fission-fetcher, and the fission-fetcher ServiceAccount was granted namespace-wide get on secrets and configmaps it needs that to load function code, env vars, and config. The runtime pod's automounted token was reachable from...

Fission router exposes /fission-function/<ns>/<name> on its public listener, allowing invocation of any function without an HTTPTrigger

Summary The Fission router registers an internal-style route — /fission-function/ and /fission-function// — for every Function object, independent of whether any HTTPTrigger exists for that function. The route was mounted on the same listener as user-defined HTTPTriggers svc/router, port 8888, so...

Fission StorageSvc /v1/archive endpoint exposes unauthenticated CRUD over all function archives

Summary The Fission storagesvc component registers archive CRUD handlers /v1/archive GET / POST / DELETE and /v1/archives list directly on its HTTP router without performing any authentication or authorization. Any caller able to reach the storagesvc ClusterIP — including any other workload in th...

Umbraco.Cms: Open Redirect Vulnerability in Surface Controllers

Impact Some of the Surface Controllers in the CMS provide to support member related operations fail to validate redirect URLs, making Razor templates that derive 'RedirectUrl' from user-controlled query parameters vulnerable to malicious redirect attacks. Patches The issue is resolved in versions...

pyload-ng: SSRF via HTTP Redirect Bypass in parse_urls API

Summary The SSRF mitigation added in commit 33c55da for GHSA-7gvf-3w72-p2pg is incomplete. The PREREQFUNCTION-based private IP check was correctly applied to HTTPChunk download path but not to HTTPRequest used by the parseurls API. An authenticated attacker can supply a URL pointing to an...

nimiq-primitives: Panic DoS in trie chunk processing via ROOT-keyed item

Impact A remote, unauthenticated denial-of-service vulnerability in MerkleRadixTrie::putchunk allows any state-sync peer to crash any node performing state synchronization freshly joining nodes and recovering nodes. A malicious peer can respond to a RequestChunk with a ResponseChunk::Chunk whose...

nimiq-blockchain: Genesis batch set request

Impact A remote peer can crash any full node by sending a RequestBatchSet message containing the genesis block's hash. The handler calls getepochchunks which iterates backwards through macro blocks using Policy::macroblockbefore. When it reaches the genesis block number, macroblockbefore panics...

nimiq-keys: Denial of service in Ed25519 multisig delinearization via invalid curve points

Impact A denial-of-service vulnerability exists in the Ed25519 multisig delinearization code path. Ed25519PublicKey::delinearize in keys/src/multisig/mod.rs called .unwrap on curve point decompression, which panics when a public key is constructed from 32 bytes that do not represent a valid point...

nimiq-primitives: BlockInclusionProof interlink issue when hops are empty

Impact A logic flaw in BlockInclusionProof::isblockproven causes the function to return true without performing any cryptographic verification when getinterlinkhops yields an empty hop list. This occurs when the target block is at the election block position immediately preceding the election...

lmdeploy: Hardcoded trust_remote_code=True is an implicit unsafe remote-code load path with no user opt-out

📋 Reframing 2026-05-02: implicit unsafe remote-code path, not "supply-chain" The accurate description of this vulnerability is: "getmodelarch and related helpers hardcode trustremotecode=True with no opt-out, creating an implicit unsafe remote-code load path on every model fetch." What this repor...

Crawlee for Python: SSRF via sitemap-derived URLs

Overview - Vulnerability type: Blind SSRF - Affected components: src/crawlee/utils/sitemap.py, src/crawlee/utils/robots.py, src/crawlee/requestloaders/sitemaprequestloader.py, and all built-in HTTP clients. - Trigger: an attacker-controlled sitemap or robots.txt containing a URL that points to an...

Apache Fory PyFory Deserialization of Untrusted Data

Fory PyFory's ReduceSerializer could bypass documented DeserializationPolicy validation hooks during reduce-state restoration and global-name resolution. An application is vulnerable if it deserializes attacker-controlled data using PyFory Python-native mode with strict mode disabled and relies o...

@sveltejs/kit: `query.batch` cross-talk

query.batch could, under very rare and specific timings, cause concurrent requests from different users to merge and resolve under single request context, enabling cross-user data disclosure...

md-fileserver: Stored/Reflected XSS when viewing Markdown (raw HTML allowed)

Summary A cross-site scripting XSS vulnerability exists in the application’s Markdown rendering logic. When user-supplied Markdown content is rendered, embedded raw HTML—including tags—is processed and injected into the resulting page without sanitization, allowing arbitrary JavaScript execution ...

Amazon SageMaker Python SDK is missing integrity verification in its Triton inference handler

Summary Amazon SageMaker Python SDK is an open-source library for training and deploying machine learning models on Amazon SageMaker. An issue exists where, under certain circumstances, the Triton inference handler deserializes model artifacts without performing integrity verification, allowing...

Cleartext storage of HMAC signing key in Amazon SageMaker Python SDK ModelBuilder/Serve path

Summary Amazon SageMaker Python SDK is an open-source library for training and deploying machine learning models on Amazon SageMaker. An issue exists where, under certain circumstances, the ModelBuilder/Serve component stores an HMAC signing key in cleartext as a container environment variable,...

LMDeploy: Arbitrary code execution via hardcoded trust_remote_code=True in lmdeploy model initialization

Summary lmdeploy hardcodes trustremotecode=True in multiple HuggingFace model-loading call sites. The affected code paths are in: text lmdeploy/archs.py lmdeploy/utils.py The vulnerable call sites pass trustremotecode=True into HuggingFace Transformers APIs such as AutoConfig.frompretrained,...

samlify: XML Injection in AttributeValue Allows Privilege Escalation in Signed SAML Assertions

Summary samlify’s template substitution only escapes attribute contexts. Values inserted into element text e.g., are not escaped. A normal user can inject XML markup into an attribute value e.g., email, name and add new elements inside the signed assertion. The IdP then signs the tampered asserti...