fix(provider): length-bound and sanitise request inputs across endpoints

.changeset/provider-input-sanitisation.md

@@ -0,0 +1,13 @@
+---
+"@prosopo/types": patch
+"@prosopo/provider": patch
+---
+
+fix(provider): length-bound and sanitise request inputs across the provider API endpoints.
+
+- Add shared zod helpers in `@prosopo/types` (`INPUT_LIMITS`, `boundedString`, `safeText`, `safeLine`): every request string field is now length-bounded, and human freetext additionally rejects control characters (null bytes etc.). Typing fields as strings already blocks Mongo operator injection; the control-character rejection covers the remaining log/header-injection vectors.
+- Apply the helpers across the provider request schemas (image/pow/puzzle captcha challenge & solution bodies, frictionless challenge, server verify, DNS event ingestion, sitekey register/remove, detector-key and decision-machine admin bodies, and the spam-email check). Tokens, signatures, behavioural/simd readings and decision-machine source get generous caps; accounts/site-keys/hashes/ids get tight ones.
+
+- Lower the provider API body-parser cap from 50 MB to 1 MB (`express.json` in `startProviderApi.ts`) as a coarse oversized-payload backstop before parsing.
+
+Email and IP fields are treated as length-bounded strings (email keeps its existing format check where present).

packages/provider/src/api/captcha/checkSpamEmail.ts

@@ -13,18 +13,19 @@
 // limitations under the License.
 
 import { ProsopoApiError } from "@prosopo/common";
+import { INPUT_LIMITS, boundedString } from "@prosopo/types";
 import type { ProviderEnvironment } from "@prosopo/types-env";
 import { extractDomainFromEmail } from "@prosopo/util";
 import type { NextFunction, Request, Response } from "express";
-import { object, string } from "zod";
+import { object } from "zod";
 import type { AugmentedRequest } from "../../express.js";
 import { Tasks } from "../../tasks/index.js";
 import { checkSpamEmail as checkSpamEmailFn } from "../../tasks/spam/checkSpamEmail.js";
 import { getMaintenanceMode } from "../admin/apiToggleMaintenanceModeEndpoint.js";
 
 const CheckSpamEmailRequestBody = object({
-	email: string(),
-	dapp: string(),
+	email: boundedString(INPUT_LIMITS.EMAIL),
+	dapp: boundedString(INPUT_LIMITS.ID),
 });
 
 export default (env: ProviderEnvironment) =>

packages/provider/src/api/startProviderApi.ts

@@ -201,7 +201,11 @@ export async function startProviderApi(
 			maxAge: 86400,
 		}),
 	);
-	apiApp.use(express.json({ limit: "50mb" }));
+	// Coarse request body-size backstop. Generous enough for legitimate
+	// payloads (captcha solutions, behavioural/simd readings, DNS event
+	// batches) but bounds oversized-payload abuse before parsing; the
+	// per-field caps in @prosopo/types (`INPUT_LIMITS`) are the finer control.
+	apiApp.use(express.json({ limit: "1mb" }));
 
 	// Put this first so that no middleware runs on it
 	apiApp.use(publicRouter(env));

packages/types/src/api/index.ts

@@ -14,3 +14,4 @@
 export * from "./api.js";
 export * from "./params.js";
 export * from "./ipapi.js";
+export * from "./sanitise.js";

packages/types/src/api/sanitise.ts

@@ -0,0 +1,82 @@
+// Copyright 2021-2026 Prosopo (UK) Ltd.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+import { string } from "zod";
+
+/**
+ * Centralised input length limits for request payloads. Generous enough not to
+ * reject legitimate input, but bounded so an oversized field cannot bloat
+ * storage, logs, downstream API calls, or act as a cheap DoS vector. The
+ * express body-size cap (provider startProviderApi.ts) is the coarse backstop;
+ * these are the per-field limits.
+ */
+export const INPUT_LIMITS = {
+	/** Identifiers, keys, slugs (accounts, site keys, dataset ids, …). */
+	ID: 256,
+	/** Names, labels, titles. */
+	NAME: 256,
+	/** Email addresses (treated as opaque strings — no format validation). */
+	EMAIL: 320,
+	/** URLs. */
+	URL: 2048,
+	/** General short freetext. Default for `boundedString` / `safeText`. */
+	TEXT: 16384,
+	/** Longer freetext: messages, descriptions, decision-machine source. */
+	LONG_TEXT: 65536,
+	/** Tokens, signatures, base64 payloads, behavioural/simd readings. */
+	TOKEN: 131072,
+} as const;
+
+// Anchored negated character classes: a string is valid only if it contains
+// NONE of these. Implemented as a `.regex()` (rather than `.refine()`) so the
+// result stays a `ZodString` and callers can still chain `.min()` / `.max()`.
+// C0 control chars + DEL (U+007F) are rejected; tab/newline/carriage-return are
+// allowed in safeText. C1 (U+0080–U+009F) is intentionally not rejected, to
+// avoid false positives on legitimate international text.
+// biome-ignore lint/suspicious/noControlCharactersInRegex: deliberately matching control chars in order to reject them
+const NO_CONTROL_CHARS = /^[^\u0000-\u0008\u000b\u000c\u000e-\u001f\u007f]*$/;
+// Single-line variant: additionally rejects tab/CR/LF.
+// biome-ignore lint/suspicious/noControlCharactersInRegex: deliberately matching control chars in order to reject them
+const NO_CONTROL_OR_NEWLINE = /^[^\u0000-\u001f\u007f]*$/;
+
+/**
+ * A length-bounded string. Use for structured values (ids, keys, tokens) where
+ * the character set is already constrained by format.
+ */
+export const boundedString = (max: number = INPUT_LIMITS.TEXT) =>
+	string().max(max);
+
+/**
+ * Length-bounded freetext that rejects control characters (null bytes etc.).
+ * Use for human-entered, multi-line text (messages, descriptions). Typing the
+ * field as a string already blocks Mongo operator injection (an object such as
+ * `{$gt:…}` fails the string check); this additionally rejects the control
+ * characters used for log/terminal injection.
+ */
+export const safeText = (max: number = INPUT_LIMITS.TEXT) =>
+	string()
+		.max(max)
+		.regex(NO_CONTROL_CHARS, "must not contain control characters");
+
+/**
+ * Single-line variant of {@link safeText}: also rejects line breaks. Use for
+ * short freetext that flows into headers/subjects (names, titles) to prevent
+ * header injection.
+ */
+export const safeLine = (max: number = INPUT_LIMITS.NAME) =>
+	string()
+		.max(max)
+		.regex(
+			NO_CONTROL_OR_NEWLINE,
+			"must not contain control characters or line breaks",
+		);