feat(voice)!: add new encryption methods, remove old methods (#10451)

BREAKING CHANGE: This library no longer supports using `tweetnacl` as an encryption library due to Discord deprecating the algorithms that `tweetnacl` helped us support (read more [here](https://discord.com/developers/docs/change-log#voice-encryption-modes)). Please migrate to one of: `sodium-native`, `sodium`, `@stablelib/xchacha20poly1305`, `@noble/ciphers` or `libsodium-wrappers` unless your system supports `aes-256-gcm` (verify by running `require('node:crypto').getCiphers().includes('aes-256-gcm')`).

---------

Co-authored-by: Jiralite <[email protected]>
Co-authored-by: Vlad Frangu <[email protected]>

Author: 3 people

packages/voice/README.md

@@ -53,7 +53,8 @@ try installing another.
 
 - `sodium-native`: ^3.3.0
 - `sodium`: ^3.0.2
-- `tweetnacl`: ^1.0.3
+- `@stablelib/xchacha20poly1305`: ^2.0.0
+- `@noble/ciphers`: ^1.0.0
 - `libsodium-wrappers`: ^0.7.9
 
 **Opus Libraries (npm install):**

packages/voice/__mocks__/rtp.ts

@@ -1,13 +1,17 @@
 import { Buffer } from 'node:buffer';
 
+// The following constants are silence packets collected from various platforms because Discord did not previously send header extensions
+// The header extension (extra data in decrypted vs opusFrame) can be detected in the position of {encrypted.subarray(12,14)} if it is equal to 0xbe,0xde
+// The header extension length will then follow as an integer and can be removed from the decrypted data (see ../src/receive/VoiceReceiver.ts:parsePacket)
+
 export const RTP_PACKET_DESKTOP = {
 	ssrc: 341_124,
 	packet: Buffer.from([
-		0x90, 0x78, 0x27, 0xe9, 0xf7, 0xcb, 0xbc, 0xd1, 0x0, 0x5, 0x34, 0x84, 0x8a, 0xbb, 0xe2, 0x97, 0x21, 0x9f, 0x1f,
-		0x67, 0xcd, 0x17, 0x91, 0x56, 0x43, 0xa0, 0x98, 0xfd, 0xa9, 0x25, 0x81, 0x63, 0x13, 0xb4, 0x1e, 0xae, 0x88, 0xe4,
-		0x0, 0xed, 0x0, 0x0, 0x0,
+		0x90, 0x78, 0x27, 0xe9, 0xf7, 0xcb, 0xbc, 0xd1, 0x0, 0x5, 0x34, 0x84, 0xbe, 0xde, 0x0, 0x1, 0x8a, 0xbb, 0xe2, 0x97,
+		0x21, 0x9f, 0x1f, 0x67, 0xcd, 0x17, 0x91, 0x56, 0x43, 0xa0, 0x98, 0xfd, 0xa9, 0x25, 0x81, 0x63, 0x13, 0xb4, 0x1e,
+		0xae, 0x88, 0xe4, 0x0, 0xed, 0x0, 0x0, 0x0,
 	]),
-	decrypted: Buffer.from([0xbe, 0xde, 0x0, 0x1, 0x10, 0xff, 0x90, 0x0, 0xf8, 0xff, 0xfe]),
+	decrypted: Buffer.from([0x10, 0xff, 0x90, 0x0, 0xf8, 0xff, 0xfe]),
 	opusFrame: Buffer.from([0xf8, 0xff, 0xfe]),
 };
 
@@ -24,10 +28,43 @@ export const RTP_PACKET_CHROME = {
 export const RTP_PACKET_ANDROID = {
 	ssrc: 172_596,
 	packet: Buffer.from([
-		0x90, 0x78, 0x39, 0xd0, 0xe0, 0x59, 0xf5, 0x47, 0x0, 0x2, 0xa2, 0x34, 0x12, 0x6d, 0x87, 0x56, 0x25, 0xc8, 0x3e,
-		0x96, 0xc0, 0x71, 0x9a, 0x1, 0x83, 0xe, 0x1, 0x62, 0x91, 0x95, 0x1f, 0x76, 0x57, 0x15, 0x41, 0xab, 0xee, 0x5b, 0xac,
-		0x8b, 0x0, 0x0, 0x0,
+		0x90, 0x78, 0x39, 0xd0, 0xe0, 0x59, 0xf5, 0x47, 0x0, 0x2, 0xa2, 0x34, 0xbe, 0xde, 0x0, 0x1, 0x12, 0x6d, 0x87, 0x56,
+		0x25, 0xc8, 0x3e, 0x96, 0xc0, 0x71, 0x9a, 0x1, 0x83, 0xe, 0x1, 0x62, 0x91, 0x95, 0x1f, 0x76, 0x57, 0x15, 0x41, 0xab,
+		0xee, 0x5b, 0xac, 0x8b, 0x0, 0x0, 0x0,
 	]),
-	decrypted: Buffer.from([0xbe, 0xde, 0x0, 0x1, 0x10, 0xff, 0x90, 0x0, 0xf8, 0xff, 0xfe]),
+	decrypted: Buffer.from([0x10, 0xff, 0x90, 0x0, 0xf8, 0xff, 0xfe]),
 	opusFrame: Buffer.from([0xf8, 0xff, 0xfe]),
 };
+
+export const XCHACHA20_SAMPLE = {
+	encrypted: Buffer.from([
+		144, 120, 86, 102, 191, 243, 52, 48, 0, 0, 191, 25, 190, 222, 0, 2, 211, 29, 103, 76, 73, 102, 37, 162, 65, 41, 39,
+		252, 26, 85, 90, 228, 241, 169, 112, 65, 95, 183, 236, 4, 87, 207, 215, 195, 180, 39, 60, 224, 221, 89, 32, 187,
+		208, 228, 145, 252, 132, 189, 103, 208, 180, 183, 217, 50, 70, 86, 20, 28, 142, 66, 251, 122, 214, 17, 7, 173, 203,
+		117, 148, 232, 118, 103, 8, 136, 227, 136, 249, 243, 135, 41, 31, 103, 146, 15, 1, 0, 0,
+	]),
+	key: new Uint8Array([
+		105, 217, 109, 27, 247, 101, 71, 49, 71, 151, 172, 85, 91, 11, 201, 23, 43, 242, 147, 81, 96, 60, 157, 50, 63, 200,
+		133, 174, 108, 144, 251, 110,
+	]),
+
+	decrypted: Buffer.from([
+		0x32, 0x64, 0xe6, 0x62, 0x10, 0xe3, 0x90, 0x02, 0x78, 0x07, 0xd6, 0x2f, 0x52, 0x23, 0x20, 0x9a, 0xab, 0x2c, 0xcc,
+		0x1c, 0x88, 0x8e, 0xcb, 0xd9, 0x4d, 0xe5, 0x33, 0x7a, 0x4b, 0x2b, 0xed, 0xa7, 0xaf, 0x5f, 0x8d, 0xb2, 0x59, 0x99,
+		0x75, 0x36, 0xf2, 0x88, 0xf5, 0xc7, 0x9f, 0x47, 0xaf, 0x92, 0x5a, 0x96, 0x3b, 0xd8, 0x9f, 0x3a, 0xb4, 0x13, 0xce,
+		0x2f, 0xae, 0x0a, 0x37,
+	]),
+};
+
+export const AES256GCM_SAMPLE = {
+	encrypted: Buffer.from([
+		128, 120, 163, 156, 159, 11, 131, 240, 0, 0, 197, 183, 175, 91, 102, 101, 195, 6, 200, 143, 117, 72, 108, 44, 165,
+		123, 121, 49, 111, 38, 3, 0, 0, 0, 90,
+	]),
+	key: Buffer.from([
+		109, 77, 195, 27, 111, 50, 231, 84, 179, 255, 217, 217, 34, 227, 19, 106, 195, 20, 150, 237, 38, 4, 101, 210, 5, 90,
+		8, 241, 58, 223, 24, 24,
+	]),
+
+	decrypted: Buffer.from([0xf8, 0xff, 0xfe]),
+};

packages/voice/__tests__/Secretbox.test.ts

@@ -1,9 +1,12 @@
 import { test, expect, vitest } from 'vitest';
-import { methods } from '../src/util/Secretbox';
+import { methods, secretboxLoadPromise } from '../src/util/Secretbox';
 
-vitest.mock('tweetnacl');
+vitest.mock('@noble/ciphers/chacha');
 
-test('Does not throw error with a package installed', () => {
-	// @ts-expect-error We are testing
-	expect(() => methods.open()).toThrow(TypeError);
+// TODO: what is this even testing exactly?
+test.skip('Does not throw error with a package installed', async () => {
+	// The async loop in Secretbox will not have finished importing unless we wait
+	await secretboxLoadPromise;
+
+	expect(() => methods.crypto_aead_xchacha20poly1305_ietf_decrypt()).not.toThrowError();
 });

packages/voice/__tests__/VoiceReceiver.test.ts

@@ -6,7 +6,13 @@ import { once } from 'node:events';
 import process from 'node:process';
 import { VoiceOpcodes } from 'discord-api-types/voice/v4';
 import { describe, test, expect, vitest, beforeEach } from 'vitest';
-import { RTP_PACKET_DESKTOP, RTP_PACKET_CHROME, RTP_PACKET_ANDROID } from '../__mocks__/rtp';
+import {
+	RTP_PACKET_DESKTOP,
+	RTP_PACKET_CHROME,
+	RTP_PACKET_ANDROID,
+	XCHACHA20_SAMPLE,
+	AES256GCM_SAMPLE,
+} from '../__mocks__/rtp';
 import { VoiceConnection, VoiceConnectionStatus } from '../src/VoiceConnection';
 import { VoiceReceiver } from '../src/receive/VoiceReceiver';
 import { methods } from '../src/util/Secretbox';
@@ -22,10 +28,6 @@ vitest.mock('../src/VoiceConnection', async (importOriginal) => {
 
 vitest.mock('../src/receive/SSRCMap');
 
-const openSpy = vitest.spyOn(methods, 'open');
-
-openSpy.mockImplementation((buffer) => buffer);
-
 async function nextTick() {
 	// eslint-disable-next-line no-promise-executor-return
 	return new Promise((resolve) => process.nextTick(resolve));
@@ -62,7 +64,7 @@ describe('VoiceReceiver', () => {
 		['RTP Packet Desktop', RTP_PACKET_DESKTOP],
 		['RTP Packet Chrome', RTP_PACKET_CHROME],
 		['RTP Packet Android', RTP_PACKET_ANDROID],
-	])('onUdpMessage: %s', async (testName, RTP_PACKET) => {
+	])('onUdpMessage: decrypt from %s', async (testName, RTP_PACKET) => {
 		receiver['decrypt'] = vitest.fn().mockImplementationOnce(() => RTP_PACKET.decrypted);
 
 		const spy = vitest.spyOn(receiver.ssrcMap, 'get');
@@ -174,47 +176,42 @@ describe('VoiceReceiver', () => {
 	describe('decrypt', () => {
 		const secretKey = new Uint8Array([1, 2, 3, 4]);
 
-		beforeEach(() => {
-			openSpy.mockClear();
-		});
+		test('decrypt: aead_xchacha20_poly1305_rtpsize', () => {
+			const nonceSpace = Buffer.alloc(24);
 
-		test('decrypt: xsalsa20_poly1305_lite', () => {
-			// Arrange
-			const buffer = range(1, 32);
-			const nonce = Buffer.alloc(4);
+			const decrypted = receiver['decrypt'](
+				XCHACHA20_SAMPLE.encrypted,
+				'aead_xchacha20_poly1305_rtpsize',
+				nonceSpace,
+				XCHACHA20_SAMPLE.key,
+			);
 
-			// Act
-			const decrypted = receiver['decrypt'](buffer, 'xsalsa20_poly1305_lite', nonce, secretKey);
+			const expectedNonce = Buffer.concat([
+				XCHACHA20_SAMPLE.encrypted.slice(XCHACHA20_SAMPLE.encrypted.length - 4),
+				Buffer.alloc(20),
+			]);
 
-			// Assert
-			expect(nonce.equals(range(29, 32))).toEqual(true);
-			expect(decrypted!.equals(range(13, 28))).toEqual(true);
+			expect(nonceSpace.equals(expectedNonce)).toEqual(true);
+			expect(decrypted.equals(XCHACHA20_SAMPLE.decrypted)).toEqual(true);
 		});
 
-		test('decrypt: xsalsa20_poly1305_suffix', () => {
-			// Arrange
-			const buffer = range(1, 64);
-			const nonce = Buffer.alloc(24);
-
-			// Act
-			const decrypted = receiver['decrypt'](buffer, 'xsalsa20_poly1305_suffix', nonce, secretKey);
-
-			// Assert
-			expect(nonce.equals(range(41, 64))).toEqual(true);
-			expect(decrypted!.equals(range(13, 40))).toEqual(true);
-		});
+		test('decrypt: aead_aes256gcm_rtpsize', () => {
+			const nonceSpace = Buffer.alloc(12);
 
-		test('decrypt: xsalsa20_poly1305', () => {
-			// Arrange
-			const buffer = range(1, 64);
-			const nonce = Buffer.alloc(12);
+			const decrypted = receiver['decrypt'](
+				AES256GCM_SAMPLE.encrypted,
+				'aead_aes256_gcm_rtpsize',
+				nonceSpace,
+				AES256GCM_SAMPLE.key,
+			);
 
-			// Act
-			const decrypted = receiver['decrypt'](buffer, 'xsalsa20_poly1305', nonce, secretKey);
+			const expectedNonce = Buffer.concat([
+				AES256GCM_SAMPLE.encrypted.subarray(AES256GCM_SAMPLE.encrypted.length - 4),
+				Buffer.alloc(8),
+			]);
 
-			// Assert
-			expect(nonce.equals(range(1, 12))).toEqual(true);
-			expect(decrypted!.equals(range(13, 64))).toEqual(true);
+			expect(nonceSpace.equals(expectedNonce)).toEqual(true);
+			expect(decrypted.equals(AES256GCM_SAMPLE.decrypted)).toEqual(true);
 		});
 	});
 });

packages/voice/package.json

@@ -74,6 +74,7 @@
 		"@discordjs/opus": "^0.9.0",
 		"@discordjs/scripts": "workspace:^",
 		"@favware/cliff-jumper": "^4.1.0",
+		"@noble/ciphers": "^1.0.0",
 		"@types/node": "18.19.45",
 		"@vitest/coverage-v8": "2.0.5",
 		"cross-env": "^7.0.3",
@@ -84,7 +85,6 @@
 		"prettier": "^3.3.3",
 		"tsup": "^8.2.4",
 		"turbo": "^2.0.14",
-		"tweetnacl": "^1.0.3",
 		"typescript": "~5.5.4",
 		"vitest": "^2.0.5",
 		"vitest-websocket-mock": "^0.3.0"

packages/voice/scripts/postbuild.mjs

@@ -3,5 +3,7 @@ import { readFile, writeFile } from 'node:fs/promises';
 const data = await readFile('./dist/index.mjs', 'utf8');
 await writeFile(
 	'./dist/index.mjs',
-	`import{createRequire as topLevelCreateRequire}from"module";const require=topLevelCreateRequire(import.meta.url);${data}`,
+	`import { createRequire as topLevelCreateRequire } from "module";
+const require = topLevelCreateRequire(import.meta.url);
+${data}`,
 );

packages/voice/src/networking/Networking.ts

@@ -2,6 +2,7 @@
 /* eslint-disable id-length */
 /* eslint-disable @typescript-eslint/unbound-method */
 import { Buffer } from 'node:buffer';
+import crypto from 'node:crypto';
 import { EventEmitter } from 'node:events';
 import { VoiceOpcodes } from 'discord-api-types/voice/v4';
 import type { CloseEvent } from 'ws';
@@ -15,7 +16,12 @@ const CHANNELS = 2;
 const TIMESTAMP_INC = (48_000 / 100) * CHANNELS;
 const MAX_NONCE_SIZE = 2 ** 32 - 1;
 
-export const SUPPORTED_ENCRYPTION_MODES = ['xsalsa20_poly1305_lite', 'xsalsa20_poly1305_suffix', 'xsalsa20_poly1305'];
+export const SUPPORTED_ENCRYPTION_MODES = ['aead_xchacha20_poly1305_rtpsize'];
+
+// Just in case there's some system that doesn't come with aes-256-gcm, conditionally add it as supported
+if (crypto.getCiphers().includes('aes-256-gcm')) {
+	SUPPORTED_ENCRYPTION_MODES.unshift('aead_aes256_gcm_rtpsize');
+}
 
 /**
  * The different statuses that a networking instance can hold. The order
@@ -187,6 +193,7 @@ function stringifyState(state: NetworkingState) {
 function chooseEncryptionMode(options: string[]): string {
 	const option = options.find((option) => SUPPORTED_ENCRYPTION_MODES.includes(option));
 	if (!option) {
+		// This should only ever happen if the gateway does not give us any encryption modes we support.
 		throw new Error(`No compatible encryption modes. Available include: ${options.join(', ')}`);
 	}
 
@@ -442,7 +449,7 @@ export class Networking extends EventEmitter {
 					sequence: randomNBit(16),
 					timestamp: randomNBit(32),
 					nonce: 0,
-					nonceBuffer: Buffer.alloc(24),
+					nonceBuffer: encryptionMode === 'aead_aes256_gcm_rtpsize' ? Buffer.alloc(12) : Buffer.alloc(24),
 					speaking: false,
 					packetsPlayed: 0,
 				},
@@ -554,18 +561,18 @@ export class Networking extends EventEmitter {
 	 * @param connectionData - The current connection data of the instance
 	 */
 	private createAudioPacket(opusPacket: Buffer, connectionData: ConnectionData) {
-		const packetBuffer = Buffer.alloc(12);
-		packetBuffer[0] = 0x80;
-		packetBuffer[1] = 0x78;
+		const rtpHeader = Buffer.alloc(12);
+		rtpHeader[0] = 0x80;
+		rtpHeader[1] = 0x78;
 
 		const { sequence, timestamp, ssrc } = connectionData;
 
-		packetBuffer.writeUIntBE(sequence, 2, 2);
-		packetBuffer.writeUIntBE(timestamp, 4, 4);
-		packetBuffer.writeUIntBE(ssrc, 8, 4);
+		rtpHeader.writeUIntBE(sequence, 2, 2);
+		rtpHeader.writeUIntBE(timestamp, 4, 4);
+		rtpHeader.writeUIntBE(ssrc, 8, 4);
 
-		packetBuffer.copy(nonce, 0, 0, 12);
-		return Buffer.concat([packetBuffer, ...this.encryptOpusPacket(opusPacket, connectionData)]);
+		rtpHeader.copy(nonce, 0, 0, 12);
+		return Buffer.concat([rtpHeader, ...this.encryptOpusPacket(opusPacket, connectionData, rtpHeader)]);
 	}
 
 	/**
@@ -574,22 +581,43 @@ export class Networking extends EventEmitter {
 	 * @param opusPacket - The Opus packet to encrypt
 	 * @param connectionData - The current connection data of the instance
 	 */
-	private encryptOpusPacket(opusPacket: Buffer, connectionData: ConnectionData) {
+	private encryptOpusPacket(opusPacket: Buffer, connectionData: ConnectionData, additionalData: Buffer) {
 		const { secretKey, encryptionMode } = connectionData;
 
-		if (encryptionMode === 'xsalsa20_poly1305_lite') {
-			connectionData.nonce++;
-			if (connectionData.nonce > MAX_NONCE_SIZE) connectionData.nonce = 0;
-			connectionData.nonceBuffer.writeUInt32BE(connectionData.nonce, 0);
-			return [
-				secretbox.methods.close(opusPacket, connectionData.nonceBuffer, secretKey),
-				connectionData.nonceBuffer.slice(0, 4),
-			];
-		} else if (encryptionMode === 'xsalsa20_poly1305_suffix') {
-			const random = secretbox.methods.random(24, connectionData.nonceBuffer);
-			return [secretbox.methods.close(opusPacket, random, secretKey), random];
-		}
+		// Both supported encryption methods want the nonce to be an incremental integer
+		connectionData.nonce++;
+		if (connectionData.nonce > MAX_NONCE_SIZE) connectionData.nonce = 0;
+		connectionData.nonceBuffer.writeUInt32BE(connectionData.nonce, 0);
+
+		// 4 extra bytes of padding on the end of the encrypted packet
+		const noncePadding = connectionData.nonceBuffer.subarray(0, 4);
+
+		let encrypted;
+		switch (encryptionMode) {
+			case 'aead_aes256_gcm_rtpsize': {
+				const cipher = crypto.createCipheriv('aes-256-gcm', secretKey, connectionData.nonceBuffer);
+				cipher.setAAD(additionalData);
 
-		return [secretbox.methods.close(opusPacket, nonce, secretKey)];
+				encrypted = Buffer.concat([cipher.update(opusPacket), cipher.final(), cipher.getAuthTag()]);
+
+				return [encrypted, noncePadding];
+			}
+
+			case 'aead_xchacha20_poly1305_rtpsize': {
+				encrypted = secretbox.methods.crypto_aead_xchacha20poly1305_ietf_encrypt(
+					opusPacket,
+					additionalData,
+					connectionData.nonceBuffer,
+					secretKey,
+				);
+
+				return [encrypted, noncePadding];
+			}
+
+			default: {
+				// This should never happen. Our encryption mode is chosen from a list given to us by the gateway and checked with the ones we support.
+				throw new RangeError(`Unsupported encryption method: ${encryptionMode}`);
+			}
+		}
 	}
 }