| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| llama.cpp before b3701 contains an out-of-bounds read vulnerability in llama_decode_internal and llama_encode_internal that allows remote unauthenticated attackers to read memory by supplying unvalidated token IDs. Attackers can send integer-array prompts to the llama-server /completion endpoint with IDs outside 0 to n_vocab to read embedding memory out of bounds or crash the process. |
| llama.cpp before b5649 contains a heap out-of-bounds read vulnerability in llama_vocab::impl::print_info() that allows attackers to crash model loading via oversized special token IDs. Attackers can supply a crafted GGUF model with a short tokenizer.ggml.tokens array, causing llama-cli or llama-server to read past id_to_token and crash. |
| ### Summary
`qs.stringify` with `allowEmptyArrays: true` writes the key of an empty array without encoding it, even though keys are encoded by default. A key that contains `&` (or the configured `delimiter`), `=`, or `#` ∴ becomes query string syntax instead of data, so a caller-supplied key name can add parameters to the serialized query string, or cut it short.
### Details
In `lib/stringify.js`, the `allowEmptyArrays` branch returns `adjustedPrefix + '[]'` early, before the code that passes every other key through `encoder` and `formatter`. So `{ "x&k=v": ["1"] }` serializes as `x%26k%3Dv%5B0%5D=1`, but `{ "x&k=v": [] }` serializes as `x&k=v[]`. The early return applies at any depth (`{ "a": { "b&k=v": [] } }` gives `a[b&k=v][]`), under every `arrayFormat`, with `allowDots`, and with a custom `encoder`, which is never called for that key. With `encode: false` or `encodeValuesOnly: true`, keys are intentionally left unencoded on every path; those configurations are not affected.
#### PoC
```js
var qs = require("qs");
var query = qs.stringify({ q: "hi", "x&apiKey=evil&y": [] }, { allowEmptyArrays: true });
// "q=hi&x&apiKey=evil&y[]"
qs.parse(query);
// { q: "hi", x: "", apiKey: "evil", y: [""] }
qs.stringify({ "x#": [], z: "1" }, { allowEmptyArrays: true });
// "x#[]&z=1": in a URL, everything after "#" becomes the fragment
```
#### Fix
Applied in 234d505 on `main` and released as v6.17.0: the key of an empty array goes through `encoder` and `formatter` exactly as the key of any other value does, so `{ "x&apiKey=evil&y": [] }` now serializes as `x%26apiKey%3Devil%26y[]`. Keys that need no encoding serialize as before, with two exceptions that match every other path: the brackets of a nested key are now encoded (`a[b][]` becomes `a%5Bb%5D[]`, which `qs.parse` reads identically), and with `encodeDotInKeys`, a dot in the key becomes `%252E` instead of `%2E` (which `decodeDotInKeys` previously misread as nesting).
### Affected versions
`>=6.12.0 <6.17.0`, fixed in v6.17.0.
`allowEmptyArrays` was introduced in f22b2bc and first shipped in v6.12.0, with this early return. Every release from v6.12.0 through v6.16.1 is affected. Earlier versions have no `allowEmptyArrays` option and are not affected.
### Impact
An application that passes an object whose key names come from untrusted input to `qs.stringify` with `allowEmptyArrays: true` (for example, to forward a parsed JSON body or user-selected fields to an upstream service) can have parameters injected into the resulting query string. The injected parameters are not keys of the object, so deleting or overwriting a key on the object (such as setting a trusted `apiKey` after copying the user's fields) does not remove them; whether the attacker's value or the application's wins depends on how the receiving side handles a duplicate parameter. A `#` in such a key ends the query string early when the result is placed in a URL. `allowEmptyArrays` is off by default, and only keys whose value (after any `filter`) is an empty array are affected. |
| The TiendaCrypto para WooCommerce WordPress plugin through 1.2.1 does not correctly verify the authenticity of its payment webhook, validating the request signature against a key that is always empty, which allows unauthenticated attackers to forge webhook notifications and mark arbitrary WooCommerce orders as paid on web servers that preserve request header case. |
| The PayDunya WooCommerce Payment Gateway WordPress plugin through 1.1.6 does not verify the authenticity of incoming payment notifications and, until the gateway is configured, authenticates them with a value derived from an empty secret that anyone can compute, allowing unauthenticated attackers to mark arbitrary WooCommerce orders as paid and completed. |
| The Kirki plugin for WordPress is vulnerable to Privilege Escalation via PHP Object Injection in all versions up to, and including, 6.2.5 via deserialization of untrusted input through the 'get-single-symbol' branch of the kirki_post_apis() AJAX dispatcher due to a missing capability check. This makes it possible for authenticated attackers, with Subscriber-level access and above, to inject a PHP Object. The additional presence of a POP chain allows attackers to overwrite an administrator's password and gain administrative access. |
| Vulnerability in the PeopleSoft Enterprise PeopleTools product of Oracle PeopleSoft (component: Fluid Core). Supported versions that are affected are 8.61-8.63. Easily exploitable vulnerability allows unauthenticated attacker with network access via HTTP to compromise PeopleSoft Enterprise PeopleTools. Successful attacks require human interaction from a person other than the attacker and while the vulnerability is in PeopleSoft Enterprise PeopleTools, attacks may significantly impact additional products (scope change). Successful attacks of this vulnerability can result in unauthorized update, insert or delete access to some of PeopleSoft Enterprise PeopleTools accessible data as well as unauthorized read access to a subset of PeopleSoft Enterprise PeopleTools accessible data. CVSS 3.1 Base Score 6.1 (Confidentiality and Integrity impacts). CVSS Vector: (CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:N). |
| Vulnerability in the Oracle Identity Manager product of Oracle Fusion Middleware (component: Connector Framework). Supported versions that are affected are 12.2.1.4.0 and 14.1.2.1.0. Easily exploitable vulnerability allows low privileged attacker with network access via HTTP to compromise Oracle Identity Manager. Successful attacks of this vulnerability can result in unauthorized ability to cause a partial denial of service (partial DOS) of Oracle Identity Manager. CVSS 3.1 Base Score 4.3 (Availability impacts). CVSS Vector: (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:L). |
| A use-after-free issue was addressed with improved memory management. This issue is fixed in Safari 27.2, iOS 26.7.2 and iPadOS 26.7.2, iOS 27.2 and iPadOS 27.2, macOS Golden Gate 27.2, visionOS 27.2. Processing maliciously crafted web content may lead to arbitrary code execution. |
| The Windows Recovery Environment (WinRE) in Windows 10 and 11 may allow an attacker to bypass firmware security controls, including administrator-configured UEFI/BIOS passwords. An attacker with physical or administrative access to a device may be able to leverage WinRE-related boot mechanisms to circumvent firmware protections and gain unauthorized access to system resources. |
| A logic issue was addressed with improved checks. This issue is fixed in macOS Golden Gate 27.2, macOS Sequoia 15.8.2, macOS Tahoe 26.7.2. An app may bypass Gatekeeper checks. |
| The Payment Integration Wompi WordPress plugin through 4.0.1 does not verify the authenticity of incoming payment notifications on a stock install, because the callback's checksum is keyed on a merchant secret that is empty until it is manually configured, allowing an unauthenticated attacker to forge a notification and mark an arbitrary order as paid. |
| Psono Server versions 5.3.5 through 7.2.3 used a permission class that did not enforce token write scope when claiming a device code. An authenticated remote attacker holding a restricted API key-derived session could claim an attacker-controlled device code and obtain a full read-write session valid for up to 365 days, bypassing the original token's scope and lifetime restrictions. This did not bypass two-factor authentication or directly disclose plaintext secrets. |
| Psono Server CE && EE versions 4.0.33 through 7.2.3 compared the request path for the bulk-secret-read endpoint with a hard-coded unprefixed path. In deployments configured with URL_PREFIX, the comparison failed and the POST request was authorized using write scope instead of read scope. An authenticated remote attacker with an unusual read-disabled, write-enabled API key could therefore retrieve encrypted secret data and nonces through the prefixed bulk-secret-read endpoint despite lacking read permission. The default configuration without URL_PREFIX was not affected. |
| Psono Server CE && EE versions 0.1.0 through 7.2.3 accepted a user-controlled host when configuring a personal Duo device and passed it to the Duo client without an allowlist or explicit timeout. An authenticated remote attacker could cause the server to connect to selected hosts over HTTPS on port 443 and distinguish failure modes, providing an internal DNS and service-reachability oracle. Connections still required TLS certificate validation, and the flaw did not provide arbitrary protocol or port access. Unresponsive targets could also occupy a server worker until the operating-system timeout. |
| Psono Server CE & EE versions 3.1.0 through 7.2.3 did not consume a stored WebAuthn challenge after assertion verification. An unauthenticated remote attacker who possessed the user's password and one captured valid WebAuthn assertion could replay that assertion with separate login tokens while the stored challenge remained unchanged, bypassing the WebAuthn second factor. The attack required capture of a valid assertion and did not itself disclose plaintext secrets. |
| Psono Server (CE & EE) versions 1.0.0 through 7.2.3 did not enforce the direction of file-transfer credentials on file-repository operations. An authenticated remote attacker with read-only access to a file could present a download transfer to the upload endpoint, obtain a signed upload URL for an attacker-selected object key, and append a chunk to the owner's file. Exploitation could upload or overwrite repository objects and consume storage, subject to transfer size, chunk-count, and lifetime limits. |
| A security flaw has been discovered in 2K Games Sid Meier's Civilization VI 1.0.12.68. The affected element is the function loadstring of the component Lua sandbox. Performing a manipulation results in type confusion. The attack may be initiated remotely. The exploit has been released to the public and may be used for attacks. The vendor was contacted early about this disclosure. |
| This issue was addressed with improved data access restriction. This issue is fixed in iOS 26.7.2 and iPadOS 26.7.2, iOS 27.2 and iPadOS 27.2, macOS Golden Gate 27.2, macOS Sequoia 15.8.2, macOS Tahoe 26.7.2, visionOS 27.2. An app may be able to access sensitive user data. |
| Datadog dd-trace-py is the Datadog Python APM client. From 1.5.0 until 3.19.9 and 4.15.6, the default query-string obfuscation regular expression can exhibit excessive backtracking while synchronously processing attacker-controlled URL query strings for HTTP span metadata. A remote attacker can cause prolonged CPU consumption, request latency, and service unavailability in affected HTTP tracing integrations that include query strings in URL span tags. The vulnerable processing is enabled by default for those integrations. This issue is fixed in versions 3.19.9 and 4.15.6. |
