GitLab Advisory Database (GLAD)

The /v1/completions request model accepts prompt as a list of text prompts or a list of token-id prompts without any outer prompt-count bound. The serving path turns each element into a separate engine input, creates one engine generator per element, merges all generators, and allocates a response slot per prompt. An authenticated API client can therefore turn one request into an attacker-chosen number of backend subrequests before any aggregate request-count …

The run-metadata update endpoint PUT /api/v1/runs/:runId/metadata applies client-supplied "operations" by passing the attacker-controlled operation.key straight into new JSONHeroPath(operation.key).set(newMetadata, value) (packages/core/src/v3/runMetadata/operations.ts:22-23), with no prototype-pollution guard (@jsonhero/path@^1.0.21 does not reject proto/constructor/prototype). A request with key: "$.proto.polluted" sets Object.prototype.polluted in the webapp process. Because every plain object then inherits that property, it corrupts unrelated code process-wide and across tenants — including Prisma query building and the Prometheus metrics client — causing query failures, …

A client that can submit message history to a Pydantic AI UI adapter can reference arbitrary files in the application's model-provider or cloud-storage account. The server forwards the reference to the model provider, which fetches it using the server's own credentials, allowing the client to read files it should not have access to.

A client that can submit message history to a Pydantic AI UI adapter can reference arbitrary files in the application's model-provider or cloud-storage account. The server forwards the reference to the model provider, which fetches it using the server's own credentials, allowing the client to read files it should not have access to.

A missing end anchor ($) in the ClassDefinition UID validation regex allows an authenticated user with the objects permission to create a class with a malicious UID containing SQL. When a data object of that class is later loaded, Block.php concatenates the raw classId directly into a SQL query without quoting, executing the injected payload. This is an incomplete fix from commit dbe1d131e4 which added a leading ^ anchor but …

nltk.data.load() and nltk.data.find() resolve user-supplied resource names to filesystem paths using url2pathname(), which decodes percent-encoded sequences (e.g. %2e%2e to ..). Path safety checks are performed on the raw, still-encoded string before decoding occurs. An attacker supplying %2e%2e instead of .. bypasses all path validation and reads arbitrary files outside the NLTK data directory.

Vault token and secret values are exposed in thrown errors when using hashi-vault-js.

src/node/handler/ImportHandler.ts and src/node/handler/ExportHandler.ts both compute their temporary working-file paths as: const randNum = Math.floor(Math.random() * 0xFFFFFFFF); const srcFile = ${os.tmpdir()}/etherpad_export_${randNum}.html; const destFile = ${os.tmpdir()}/etherpad_export_${randNum}.${type}; Two flaws compound: Math.random() is not crypto-secure. It yields at most ~32 bits of entropy and is predictable across calls within the same Node process (V8 shares PRNG state between consecutive Math.random() invocations). An attacker on the same host who observes any earlier temp-file name from …

Insufficient clearing of the output buffer in Java-based decompressor implementations in lz4-java 1.10.0 and earlier allows remote attackers to read previous buffer contents via crafted compressed input. In applications where the output buffer is reused without being cleared, this may lead to disclosure of sensitive data. JNI-based implementations are not affected.

Insufficient clearing of the output buffer in Java-based decompressor implementations in lz4-java 1.10.0 and earlier allows remote attackers to read previous buffer contents via crafted compressed input. In applications where the output buffer is reused without being cleared, this may lead to disclosure of sensitive data. JNI-based implementations are not affected.

Insufficient clearing of the output buffer in Java-based decompressor implementations in lz4-java 1.10.0 and earlier allows remote attackers to read previous buffer contents via crafted compressed input. In applications where the output buffer is reused without being cleared, this may lead to disclosure of sensitive data. JNI-based implementations are not affected.

This advisory has been marked as a false positive.

This advisory has been marked as a false positive.

This advisory has been marked as a false positive.

This advisory has been marked as False Positive and has been removed.

After an API Token exposure from an exploited trivy dependency, two new releases of litellm were uploaded to PyPI containing automatically activated malware, harvesting sensitive credentials and files, and exfiltrating to a remote API. Anyone who has installed and run the project should assume any credentials available to litellm environment may have been exposed, and revoke/rotate thema ccordingly.