Merge pull request #220 from qeled/indev

Implement voice encryption support 0.0

lib/Voice/VoiceConnection.js

@@ -51,6 +51,9 @@ var _unpipe = require("unpipe");
 
 var _unpipe2 = _interopRequireDefault(_unpipe);
 
+var MODE_xsalsa20_poly1305 = "xsalsa20_poly1305";
+var MODE_plain = "plain";
+
 var VoiceConnection = (function (_EventEmitter) {
 	_inherits(VoiceConnection, _EventEmitter);
 
@@ -77,6 +80,10 @@ var VoiceConnection = (function (_EventEmitter) {
 		this.KAI = null;
 		this.timestamp = 0;
 		this.sequence = 0;
+
+		this.mode = null;
+		this.secret = null;
+
 		this.volume = new _VolumeTransformer2["default"]();
 		this.init();
 	}
@@ -232,7 +239,7 @@ var VoiceConnection = (function (_EventEmitter) {
 			}
 
 			var buffer = self.encoder.opusBuffer(rawbuffer);
-			var packet = new _VoicePacket2["default"](buffer, sequence, timestamp, self.vWSData.ssrc);
+			var packet = new _VoicePacket2["default"](buffer, sequence, timestamp, self.vWSData.ssrc, self.secret);
 			return self.sendPacket(packet, callback);
 		} catch (e) {
 			self.playing = false;
@@ -361,6 +368,13 @@ var VoiceConnection = (function (_EventEmitter) {
 					}
 					discordPort = msg.readUIntLE(msg.length - 2, 2).toString(10);
 
+					var modes = self.vWSData.modes;
+					var mode = MODE_xsalsa20_poly1305;
+					if (modes.indexOf(MODE_xsalsa20_poly1305) < 0) {
+						mode = MODE_plain;
+						self.client.emit("debug", "Encrypted mode not reported as supported by the server, using 'plain'");
+					}
+
 					vWS.send(JSON.stringify({
 						"op": 1,
 						"d": {
@@ -368,7 +382,7 @@ var VoiceConnection = (function (_EventEmitter) {
 							"data": {
 								"address": discordIP,
 								"port": Number(discordPort),
-								"mode": self.vWSData.modes[0] //Plain
+								"mode": mode
 							}
 						}
 					}));
@@ -410,6 +424,13 @@ var VoiceConnection = (function (_EventEmitter) {
 						});
 						break;
 					case 4:
+						if (data.d.secret_key && data.d.secret_key.length > 0) {
+							var buffer = new ArrayBuffer(data.d.secret_key.length);
+							self.secret = new Uint8Array(buffer);
+							for (var i = 0; i < _this4.secret.length; i++) {
+								self.secret[i] = data.d.secret_key[i];
+							}
+						}
 
 						self.ready = true;
 						self.mode = data.d.mode;

lib/Voice/VoicePacket.js

@@ -2,27 +2,45 @@
 
 exports.__esModule = true;
 
+function _interopRequireDefault(obj) { return obj && obj.__esModule ? obj : { "default": obj }; }
+
 function _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError("Cannot call a class as a function"); } }
 
-var VoicePacket = function VoicePacket(data, sequence, time, ssrc) {
+var _tweetnacl = require("tweetnacl");
-	_classCallCheck(this, VoicePacket);
 
-	var audioBuffer = data,
+var _tweetnacl2 = _interopRequireDefault(_tweetnacl);
-	    returnBuffer = new Buffer(audioBuffer.length + 12);
 
-	returnBuffer.fill(0);
+var nonce = new Buffer(24);
-	returnBuffer[0] = 0x80;
+nonce.fill(0);
-	returnBuffer[1] = 0x78;
 
-	returnBuffer.writeUIntBE(sequence, 2, 2);
+var VoicePacket = function VoicePacket(data, sequence, time, ssrc, secret) {
-	returnBuffer.writeUIntBE(time, 4, 4);
+		_classCallCheck(this, VoicePacket);
-	returnBuffer.writeUIntBE(ssrc, 8, 4);
 
-	for (var i = 0; i < audioBuffer.length; i++) {
+		var mac = secret ? 16 : 0;
-		returnBuffer[i + 12] = audioBuffer[i];
+		var packetLength = data.length + 12 + mac;
-	}
 
-	return returnBuffer;
+		var audioBuffer = data;
+		var returnBuffer = new Buffer(packetLength);
+
+		returnBuffer.fill(0);
+		returnBuffer[0] = 0x80;
+		returnBuffer[1] = 0x78;
+
+		returnBuffer.writeUIntBE(sequence, 2, 2);
+		returnBuffer.writeUIntBE(time, 4, 4);
+		returnBuffer.writeUIntBE(ssrc, 8, 4);
+
+		if (secret) {
+				// copy first 12 bytes
+				returnBuffer.copy(nonce, 0, 0, 12);
+				audioBuffer = _tweetnacl2["default"].secretbox(data, nonce, secret);
+		}
+
+		for (var i = 0; i < audioBuffer.length; i++) {
+				returnBuffer[i + 12] = audioBuffer[i];
+		}
+
+		return returnBuffer;
 };
 
 exports["default"] = VoicePacket;

package.json

@@ -26,6 +26,7 @@
   "homepage": "https://github.com/hydrabolt/discord.js#readme",
   "dependencies": {
     "superagent": "^1.8.0",
+    "tweetnacl": "^0.14.1",
     "unpipe": "^1.0.0",
     "ws": "^1.0.1"
   },

src/Voice/VoiceConnection.js

@@ -17,6 +17,9 @@ import StreamIntent from "./StreamIntent";
 import EventEmitter from "events";
 import unpipe from "unpipe";
 
+const MODE_xsalsa20_poly1305 = "xsalsa20_poly1305";
+const MODE_plain = "plain";
+
 export default class VoiceConnection extends EventEmitter {
 	constructor(channel, client, session, token, server, endpoint) {
 		super();
@@ -39,6 +42,10 @@ export default class VoiceConnection extends EventEmitter {
 		this.KAI = null;
 		this.timestamp = 0;
 		this.sequence = 0;
+
+		this.mode = null;
+		this.secret = null;
+
 		this.volume = new VolumeTransformer();
 		this.init();
 	}
@@ -196,7 +203,7 @@ export default class VoiceConnection extends EventEmitter {
 			}
 
 			var buffer = self.encoder.opusBuffer(rawbuffer);
-			var packet = new VoicePacket(buffer, sequence, timestamp, self.vWSData.ssrc);
+			var packet = new VoicePacket(buffer, sequence, timestamp, self.vWSData.ssrc, self.secret);
 			return self.sendPacket(packet, callback);
 
 		} catch (e) {
@@ -314,6 +321,13 @@ export default class VoiceConnection extends EventEmitter {
 					}
 					discordPort = msg.readUIntLE(msg.length - 2, 2).toString(10);
 
+					var modes = self.vWSData.modes;
+					var mode = MODE_xsalsa20_poly1305;
+					if (modes.indexOf(MODE_xsalsa20_poly1305) < 0) {
+						mode = MODE_plain;
+						self.client.emit("debug", "Encrypted mode not reported as supported by the server, using 'plain'");
+					}
+
 					vWS.send(JSON.stringify({
 						"op": 1,
 						"d": {
@@ -321,7 +335,7 @@ export default class VoiceConnection extends EventEmitter {
 							"data": {
 								"address": discordIP,
 								"port": Number(discordPort),
-								"mode": self.vWSData.modes[0] //Plain
+								"mode": mode
 							}
 						}
 					}));
@@ -365,6 +379,13 @@ export default class VoiceConnection extends EventEmitter {
 						});
 						break;
 					case 4:
+						if (data.d.secret_key && data.d.secret_key.length > 0) {
+							const buffer = new ArrayBuffer(data.d.secret_key.length);
+							self.secret = new Uint8Array(buffer);
+							for (let i = 0; i < this.secret.length; i++) {
+								self.secret[i] = data.d.secret_key[i];
+							}
+						}
 
 						self.ready = true;
 						self.mode = data.d.mode;

src/Voice/VoicePacket.js

@@ -1,10 +1,17 @@
 "use strict";
 
-export default class VoicePacket {
+import nacl from "tweetnacl";
-	constructor(data, sequence, time, ssrc){
 
-		var audioBuffer = data,
+const nonce = new Buffer(24);
-			returnBuffer = new Buffer(audioBuffer.length + 12);
+nonce.fill(0);
+
+export default class VoicePacket {
+	constructor(data, sequence, time, ssrc, secret){
+		var mac = secret ? 16 : 0;
+		var packetLength = data.length + 12 + mac;
+
+		var audioBuffer = data;
+		var returnBuffer = new Buffer(packetLength);
 
 		returnBuffer.fill(0);
 		returnBuffer[0] = 0x80;
@@ -14,7 +21,13 @@ export default class VoicePacket {
 		returnBuffer.writeUIntBE(time, 4, 4);
 		returnBuffer.writeUIntBE(ssrc, 8, 4);
 
-		for (var i=0; i<audioBuffer.length; i++) {
+		if (secret) {
+			// copy first 12 bytes
+			returnBuffer.copy(nonce, 0, 0, 12);
+			audioBuffer = nacl.secretbox(data, nonce, secret);
+		}
+
+		for (var i = 0; i < audioBuffer.length; i++) {
 			returnBuffer[i + 12] = audioBuffer[i];
 		}