fix(whatsapp): gate pairing access-control on extractable inbound user content (#73797) (#73823)

Merged via squash.

Prepared head SHA: 61506e1439
Co-authored-by: hclsys <7755017+hclsys@users.noreply.github.com>
Co-authored-by: mcaxtr <7562095+mcaxtr@users.noreply.github.com>
Reviewed-by: @mcaxtr

extensions/whatsapp/src/inbound/extract.test.ts

@@ -1,6 +1,6 @@
 import type { proto } from "@whiskeysockets/baileys";
 import { describe, expect, it } from "vitest";
-import { extractMentionedJids } from "./extract.js";
+import { extractMentionedJids, hasInboundUserContent } from "./extract.js";
 
 describe("extractMentionedJids", () => {
   const botJid = "5511999999999@s.whatsapp.net";
@@ -101,3 +101,182 @@ describe("extractMentionedJids", () => {
     expect(extractMentionedJids(message)).toEqual([botJid]);
   });
 });
+
+describe("hasInboundUserContent", () => {
+  it("returns true for plain text conversation", () => {
+    expect(hasInboundUserContent({ conversation: "hello" })).toBe(true);
+  });
+
+  it("returns true for extendedTextMessage", () => {
+    expect(
+      hasInboundUserContent({ extendedTextMessage: { text: "hello" } } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for image message", () => {
+    expect(
+      hasInboundUserContent({ imageMessage: { mimetype: "image/png" } } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for video message", () => {
+    expect(
+      hasInboundUserContent({ videoMessage: { mimetype: "video/mp4" } } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for audio message", () => {
+    expect(
+      hasInboundUserContent({ audioMessage: { mimetype: "audio/ogg" } } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for document message", () => {
+    expect(
+      hasInboundUserContent({
+        documentMessage: { fileName: "x.pdf" },
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for sticker message", () => {
+    expect(
+      hasInboundUserContent({ stickerMessage: { mimetype: "image/webp" } } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for location message with valid coords", () => {
+    expect(
+      hasInboundUserContent({
+        locationMessage: { degreesLatitude: 1, degreesLongitude: 2 },
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for live location message with valid coords", () => {
+    expect(
+      hasInboundUserContent({
+        liveLocationMessage: { degreesLatitude: 1, degreesLongitude: 2 },
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for contact message", () => {
+    expect(
+      hasInboundUserContent({
+        contactMessage: { displayName: "Alice", vcard: "BEGIN:VCARD\nEND:VCARD" },
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for contactsArrayMessage via contact placeholder extraction", () => {
+    expect(
+      hasInboundUserContent({
+        contactsArrayMessage: {
+          contacts: [{ displayName: "Alice", vcard: "BEGIN:VCARD\nEND:VCARD" }],
+        },
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for buttons response (user button click)", () => {
+    expect(
+      hasInboundUserContent({
+        buttonsResponseMessage: {
+          selectedButtonId: "yes",
+          selectedDisplayText: "Yes",
+        },
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for list response (user list selection)", () => {
+    expect(
+      hasInboundUserContent({
+        listResponseMessage: {
+          title: "Option A",
+          singleSelectReply: { selectedRowId: "a" },
+        } as unknown as proto.Message.IListResponseMessage,
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for template button reply", () => {
+    expect(
+      hasInboundUserContent({
+        templateButtonReplyMessage: {
+          selectedId: "btn-1",
+          selectedDisplayText: "Click",
+        } as unknown as proto.Message.ITemplateButtonReplyMessage,
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for interactive response", () => {
+    expect(
+      hasInboundUserContent({
+        interactiveResponseMessage: {
+          body: { text: "x" },
+          nativeFlowResponseMessage: { name: "n", paramsJson: "{}" },
+        } as unknown as proto.Message.IInteractiveResponseMessage,
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns true for buttons response wrapped in ephemeralMessage (regression for #73797 + greptile review)", () => {
+    expect(
+      hasInboundUserContent({
+        ephemeralMessage: {
+          message: {
+            buttonsResponseMessage: {
+              selectedButtonId: "ok",
+              selectedDisplayText: "OK",
+            },
+          },
+        },
+      } as proto.IMessage),
+    ).toBe(true);
+  });
+
+  it("returns false for undefined message (regression for #73797)", () => {
+    expect(hasInboundUserContent(undefined)).toBe(false);
+  });
+
+  it("returns false for empty message object (no content keys)", () => {
+    expect(hasInboundUserContent({} as proto.IMessage)).toBe(false);
+  });
+
+  it("returns false for protocol message envelope without inner content (regression for #73797)", () => {
+    expect(
+      hasInboundUserContent({
+        protocolMessage: {
+          type: 0,
+        } as unknown as proto.Message.IProtocolMessage,
+      } as proto.IMessage),
+    ).toBe(false);
+  });
+
+  it("returns false for receipt-style senderKeyDistribution-only payload (regression for #73797)", () => {
+    expect(
+      hasInboundUserContent({
+        senderKeyDistributionMessage: {
+          groupId: "g@example",
+        } as unknown as proto.Message.ISenderKeyDistributionMessage,
+      } as proto.IMessage),
+    ).toBe(false);
+  });
+
+  it("returns false when location coords are missing (incomplete event, regression for #73797)", () => {
+    expect(
+      hasInboundUserContent({
+        locationMessage: { name: "no coords" },
+      } as proto.IMessage),
+    ).toBe(false);
+  });
+
+  it("returns false when extendedTextMessage has only empty text", () => {
+    expect(hasInboundUserContent({ extendedTextMessage: { text: "  " } } as proto.IMessage)).toBe(
+      false,
+    );
+  });
+});

extensions/whatsapp/src/inbound/extract.ts

@@ -438,3 +438,49 @@ export function describeReplyContext(
     sender,
   };
 }
+
+function hasInteractiveResponseContent(message: proto.IMessage | undefined): boolean {
+  if (!message) {
+    return false;
+  }
+  // Button/list/template/interactive selections that the existing four
+  // extractors do not cover. Treat any presence of these keys as user
+  // content — Baileys never delivers these as receipts or protocol
+  // envelopes, only as explicit user choices.
+  return Boolean(
+    message.buttonsResponseMessage ||
+    message.listResponseMessage ||
+    message.templateButtonReplyMessage ||
+    message.interactiveResponseMessage,
+  );
+}
+
+/**
+ * Fast check that a Baileys message carries user-visible inbound content
+ * (text, media, contact, location, button/list selection). Returns false for
+ * protocol/receipt/typing notifications that arrive on the same
+ * `messages.upsert` stream as real messages but should not trigger pairing
+ * access-control side effects.
+ */
+export function hasInboundUserContent(rawMessage: proto.IMessage | undefined): boolean {
+  if (!rawMessage) {
+    return false;
+  }
+  if (extractText(rawMessage)) {
+    return true;
+  }
+  if (extractMediaPlaceholder(rawMessage)) {
+    return true;
+  }
+  if (extractLocationData(rawMessage)) {
+    return true;
+  }
+  // Walk wrappers (ephemeral, viewOnce, etc.) — interactive responses
+  // can arrive nested.
+  for (const candidate of buildMessageChain(rawMessage)) {
+    if (hasInteractiveResponseContent(candidate)) {
+      return true;
+    }
+  }
+  return false;
+}

extensions/whatsapp/src/inbound/monitor.ts

@@ -35,6 +35,7 @@ import {
   extractMediaPlaceholder,
   extractMentionedJids,
   extractText,
+  hasInboundUserContent,
 } from "./extract.js";
 import { attachEmitterListener, closeInboundMonitorSocket } from "./lifecycle.js";
 import { downloadInboundMedia } from "./media.js";
@@ -381,6 +382,18 @@ export async function attachWebInboxToSocket(
       );
       return null;
     }
+    // Gate pairing access-control on extractable inbound user content. Baileys
+    // delivers receipts, typing indicators, presence updates, and protocol
+    // messages on the same `messages.upsert` stream as real messages; without
+    // this gate, `checkInboundAccessControl` can send an unsolicited pairing
+    // verification reply to a `dmPolicy: pairing` peer who never typed
+    // anything (e.g. when Master sends an outbound message to a new JID and
+    // the receipt round-trip arrives before the recipient ever replies).
+    // Echoes of our own outbound messages are already handled above.
+    if (!hasInboundUserContent(msg.message ?? undefined)) {
+      return null;
+    }
+
     const participantJid = msg.key?.participant ?? undefined;
     const from = group ? remoteJid : await resolveInboundJid(remoteJid);
     if (!from) {