YouTube.js/lib/deciphers/NToken.js

'use strict';

const { NTOKEN_REGEX, BASE64_DIALECT } = require('../utils/Constants');

const NTokenTransformOperation = exports.NTokenTransformOperation = {
  NO_OP: 0,
  PUSH: 1,
  REVERSE_1: 2,
  REVERSE_2: 3,
  SPLICE: 4,
  SWAP0_1: 5,
  SWAP0_2: 6,
  ROTATE_1: 7,
  ROTATE_2: 8,
  BASE64_DIA: 9,
  TRANSLATE_1: 10,
  TRANSLATE_2: 11
};

const NTokenTransformOpType = exports.NTokenTransformOpType = {
  FUNC: 0,
  N_ARR: 1,
  LITERAL: 2,
  REF: 3
};

const OP_LOOKUP = {
  'd.push(e)': NTokenTransformOperation.PUSH,
  'd.reverse()': NTokenTransformOperation.REVERSE_1,
  'function(d){for(var': NTokenTransformOperation.REVERSE_2,
  'd.length;d.splice(e,1)': NTokenTransformOperation.SPLICE,
  'd[0])[0])': NTokenTransformOperation.SWAP0_1,
  'f=d[0];d[0]': NTokenTransformOperation.SWAP0_2,
  'reverse().forEach': NTokenTransformOperation.ROTATE_1,
  'unshift(d.pop())': NTokenTransformOperation.ROTATE_2,
  'function(){for(var': NTokenTransformOperation.BASE64_DIA,
  'function(d,e){for(var f': NTokenTransformOperation.TRANSLATE_1,
  'function(d,e,f){var': NTokenTransformOperation.TRANSLATE_2
};

class NTokenTransforms {
  /**
   * Gets a base64 alphabet and uses it as a lookup table to modify n.
   *
   * @param {any[]} arr
   * @param {string} token
   * @param {boolean} is_reverse_base64
   * @returns {void}
   */
  static translate1(arr, token, is_reverse_base64) {
    const characters = is_reverse_base64 ? BASE64_DIALECT.REVERSE : BASE64_DIALECT.NORMAL;
    const token_chars = token.split('');
    arr.forEach((char, index, loc) => {
      token_chars.push(loc[index] = characters[(characters.indexOf(char) - characters.indexOf(token_chars[index]) + 64) % characters.length]);
    });
  }

  static translate2(arr, token, characters) {
    let chars_length = characters.length;
    const token_chars = token.split('');
    arr.forEach((char, index, loc) => {
      token_chars.push(loc[index] = characters[(characters.indexOf(char) - characters.indexOf(token_chars[index]) + index + chars_length--) % characters.length]);
    });
  }

  /**
   * Returns the requested base64 dialect, currently this is only used by 'translate2'.
   *
   * @param {boolean} is_reverse_base64
   * @returns {string[]}
   */
  static getBase64Dia(is_reverse_base64) {
    const characters = is_reverse_base64 ? BASE64_DIALECT.REVERSE : BASE64_DIALECT.NORMAL;
    return characters;
  }

  /**
   * Swaps the first element with the one at the given index.
   *
   * @param {any[]} arr
   * @param {number} index
   * @returns {void}
   */
  static swap0(arr, index) {
    const old_elem = arr[0];
    index = (index % arr.length + arr.length) % arr.length;
    arr[0] = arr[index];
    arr[index] = old_elem;
  }

  /**
   * Rotates elements of the array.
   *
   * @param {any[]} arr
   * @param {number} index
   * @returns {void}
   */
  static rotate(arr, index) {
    index = (index % arr.length + arr.length) % arr.length;
    arr.splice(-index).reverse().forEach((el) => arr.unshift(el));
  }

  /**
   * Deletes one element at the given index.
   *
   * @param {any[]} arr
   * @param {number} index
   * @returns {void}
   */
  static splice(arr, index) {
    index = (index % arr.length + arr.length) % arr.length;
    arr.splice(index, 1);
  }

  static reverse(arr) {
    arr.reverse();
  }

  static push(arr, item) {
    if (Array.isArray(arr?.[0]))
      arr.push([ NTokenTransformOpType.LITERAL, item ]);
    else
      arr.push(item);
  }
}

exports.NTokenTransforms = NTokenTransforms;

const TRANSFORM_FUNCTIONS = [ {
  [NTokenTransformOperation.PUSH]: NTokenTransforms.push,
  [NTokenTransformOperation.SPLICE]: NTokenTransforms.splice,
  [NTokenTransformOperation.SWAP0_1]: NTokenTransforms.swap0,
  [NTokenTransformOperation.SWAP0_2]: NTokenTransforms.swap0,
  [NTokenTransformOperation.ROTATE_1]: NTokenTransforms.rotate,
  [NTokenTransformOperation.ROTATE_2]: NTokenTransforms.rotate,
  [NTokenTransformOperation.REVERSE_1]: NTokenTransforms.reverse,
  [NTokenTransformOperation.REVERSE_2]: NTokenTransforms.reverse,
  [NTokenTransformOperation.BASE64_DIA]: () => NTokenTransforms.getBase64Dia(false),
  [NTokenTransformOperation.TRANSLATE_1]: (...args) => NTokenTransforms.translate1.apply(null, [ ...args, false ]),
  [NTokenTransformOperation.TRANSLATE_2]: NTokenTransforms.translate2
}, {
  [NTokenTransformOperation.PUSH]: NTokenTransforms.push,
  [NTokenTransformOperation.SPLICE]: NTokenTransforms.splice,
  [NTokenTransformOperation.SWAP0_1]: NTokenTransforms.swap0,
  [NTokenTransformOperation.SWAP0_2]: NTokenTransforms.swap0,
  [NTokenTransformOperation.ROTATE_1]: NTokenTransforms.rotate,
  [NTokenTransformOperation.ROTATE_2]: NTokenTransforms.rotate,
  [NTokenTransformOperation.REVERSE_1]: NTokenTransforms.reverse,
  [NTokenTransformOperation.REVERSE_2]: NTokenTransforms.reverse,
  [NTokenTransformOperation.BASE64_DIA]: () => NTokenTransforms.getBase64Dia(true),
  [NTokenTransformOperation.TRANSLATE_1]: (...args) => NTokenTransforms.translate1.apply(null, [ ...args, true ]),
  [NTokenTransformOperation.TRANSLATE_2]: NTokenTransforms.translate2
} ];

class NToken {
  constructor(transformer) {
    this.transformer = transformer;
  }

  static fromSourceCode(raw) {
    const transformation_data = NToken.getTransformationData(raw);

    const transformations = transformation_data.map((el) => {
      if (el != null && typeof el != 'number') {
        const is_reverse_base64 = el.includes('case 65:');
        const opcode = OP_LOOKUP[NToken.getFunc(el)?.[0]];
        if (opcode) {
          el = [
            NTokenTransformOpType.FUNC,
            opcode, 0 + is_reverse_base64
          ];
        } else if (el == 'b') {
          el = [ NTokenTransformOpType.N_ARR ];
        } else {
          el = [ NTokenTransformOpType.LITERAL, el ];
        }
      } else if (el != null) {
        el = [ NTokenTransformOpType.LITERAL, el ];
      }

      return el;
    });

    // Fills all placeholders with the transformations array
    const placeholder_indexes = [ ...raw.matchAll(NTOKEN_REGEX.PLACEHOLDERS) ].map((item) => parseInt(item[1]));
    placeholder_indexes.forEach((i) => transformations[i] = [ NTokenTransformOpType.REF ]);

    // Parses and emulates calls to the functions of the transformations array
    const function_calls = [ ...(raw.replace(/\n/g, '').match(/try\{(.*?)\}catch/s)[1])
      .matchAll(NTOKEN_REGEX.CALLS) ].map((params) => [
      parseInt(params[1]),
      params[2].split(',').map((param) => parseInt(param.match(/c\[(.*?)\]/)?.[1]))
    ]);

    return new NToken([ transformations, function_calls ]);
  }

  evaluate(i, n_token, transformer) {
    switch (i[0]) {
      case NTokenTransformOpType.FUNC:
        return TRANSFORM_FUNCTIONS[i[2]][i[1]];
      case NTokenTransformOpType.N_ARR:
        return n_token;
      case NTokenTransformOpType.LITERAL:
        return i[1];
      case NTokenTransformOpType.REF:
        return transformer[0];
    }
  }

  transform(n) {
    const n_token = n.split('');

    // We must copy since we will modify the array
    const transformer = this.getTransformerClone();

    try {
      transformer[1].forEach(([ index, param_index ]) => {
        const base64_dia = (param_index[2] && this.evaluate(transformer[0][param_index[2]], n_token, transformer)());
        this.evaluate(transformer[0][index], n_token, transformer)(
          param_index[0] !== undefined &&
          this.evaluate(transformer[0][param_index[0]], n_token, transformer),
          param_index[1] !== undefined &&
          this.evaluate(transformer[0][param_index[1]], n_token, transformer),
          base64_dia
        );
      });
    } catch (err) {
      console.error(new Error(`Could not transform n-token, download may be throttled.\nOriginal Token:${n}Error:\n${err}`));
      return n;
    }

    return n_token.join('');
  }

  getTransformerClone() {
    return [
      [ ...this.transformer[0] ],
      [ ...this.transformer[1] ]
    ];
  }

  toJSON() {
    return this.getTransformerClone();
  }

  toArrayBuffer() {
    // (16 bit FUNC instructions) 2 bit op - 1 bit is_reverse_base64 - 4 bit nonce - 8 bit operation
    // (8 bit N_ARG and REF) 2 bit op - 6 bit nonce
    // (40 bit LITERAL) 2 bit op - 6 bit nonce - 32 bit value
    // NTokenCall will be 8 bit for the index, 8 bit for the number of parameters, and 8 bit for each parameter
    // We've got a 3 * 32 bit header to store the library version and the size of the two arrays

    let size = 4 * 3;

    for (const instruction of this.transformer[0]) {
      switch (instruction[0]) {
        case NTokenTransformOpType.FUNC:
          size += 2;
          break;
        case NTokenTransformOpType.N_ARR:
        case NTokenTransformOpType.REF:
          size += 1;
          break;
        case NTokenTransformOpType.LITERAL:
          if (typeof instruction[1] === 'string')
            size += 1 + 4 + new TextEncoder().encode(instruction[1]).byteLength;
          size += 4 + 1;
          break;
      }
    }

    for (const call of this.transformer[1]) {
      size += 2 + call[1].length;
    }

    const buffer = new ArrayBuffer(size);
    const view = new DataView(buffer);

    let offset = 0;

    view.setUint32(offset, NToken.LIBRARY_VERSION, true);
    offset += 4;

    view.setUint32(offset, this.transformer[0].length, true);
    offset += 4;

    view.setUint32(offset, this.transformer[1].length, true);
    offset += 4;

    for (const instruction of this.transformer[0]) {
      switch (instruction[0]) {
        case NTokenTransformOpType.FUNC: {
          const opcode = (instruction[0] << 6) | instruction[2];
          view.setUint8(offset, opcode);
          offset += 1;
          view.setUint8(offset, instruction[1]);
          offset += 1;
        }
          break;
        case NTokenTransformOpType.N_ARR:
        case NTokenTransformOpType.REF: {
          const opcode = (instruction[0] << 6);
          view.setUint8(offset, opcode);
          offset += 1;
        }
          break;
        case NTokenTransformOpType.LITERAL: {
          const type = typeof instruction[1] === 'string' ? 1 : 0;
          const opcode = (instruction[0] << 6) | type;

          view.setUint8(offset, opcode);
          offset += 1;

          if (type === 0) {
            view.setInt32(offset, instruction[1], true);
            offset += 4;
          } else {
            const encoded = new TextEncoder().encode(instruction[1]);
            view.setUint32(offset, encoded.byteLength, true);
            offset += 4;
            for (let i = 0; i < encoded.byteLength; i++) {
              view.setUint8(offset, encoded[i]);
              offset += 1;
            }
          }
        }
          break;
      }
    }

    for (const call of this.transformer[1]) {
      view.setUint8(offset, call[0]);
      offset += 1;

      view.setUint8(offset, call[1].length);
      offset += 1;

      for (const param of call[1]) {
        view.setUint8(offset, param);
        offset += 1;
      }
    }

    return buffer;
  }

  static fromArrayBuffer(buffer) {
    const view = new DataView(buffer);
    let offset = 0;

    const version = view.getUint32(offset, true);
    offset += 4;

    if (version !== NToken.LIBRARY_VERSION)
      throw new TypeError('Invalid library version');

    const transformations_length = view.getUint32(offset, true);
    offset += 4;

    const function_calls_length = view.getUint32(offset, true);
    offset += 4;

    const transformations = new Array(transformations_length);

    for (let i = 0; i < transformations_length; i++) {
      const opcode = view.getUint8(offset++);
      const op = opcode >> 6;

      switch (op) {
        case NTokenTransformOpType.FUNC: {
          const is_reverse_base64 = opcode & 0b00000001;
          const operation = view.getUint8(offset++);
          transformations[i] = [ op, operation, is_reverse_base64 ];
        }
          break;
        case NTokenTransformOpType.N_ARR:
        case NTokenTransformOpType.REF:
          transformations[i] = [ op ];
          break;
        case NTokenTransformOpType.LITERAL: {
          const type = opcode & 0b00000001;

          if (type === 0) {
            const literal = view.getInt32(offset, true);
            offset += 4;

            transformations[i] = [ op, literal ];
          } else {
            const length = view.getUint32(offset, true);
            offset += 4;

            const literal = new Uint8Array(length);

            for (let i = 0; i < length; i++) {
              literal[i] = view.getUint8(offset++);
            }

            transformations[i] = [ op, new TextDecoder().decode(literal) ];
          }
        }
          break;
        default:
          throw new Error('Invalid opcode');
      }
    }

    const function_calls = new Array(function_calls_length);

    for (let i = 0; i < function_calls_length; i++) {
      const index = view.getUint8(offset++);
      const num_params = view.getUint8(offset++);

      const params = new Array(num_params);

      for (let j = 0; j < num_params; j++) {
        params[j] = view.getUint8(offset++);
      }

      function_calls[i] = [ index, params ];
    }

    return new NToken([ transformations, function_calls ]);
  }

  static get LIBRARY_VERSION() {
    return 1;
  }

  static getFunc(el) {
    return el.match(NTOKEN_REGEX.FUNCTIONS);
  }

  static getTransformationData(raw) {
    const data = `[${raw.replace(/\n/g, '').match(/c=\[(.*?)\];c/s)?.[1]}]`;
    return JSON.parse(this.refineNTokenData(data));
  }

  // TODO: refactor this
  static refineNTokenData(data) {
    return data
      .replace(/function\(d,e\)/g, '"function(d,e)').replace(/function\(d\)/g, '"function(d)')
      .replace(/function\(\)/g, '"function()').replace(/function\(d,e,f\)/g, '"function(d,e,f)')
      .replace(/\[function\(d,e,f\)/g, '["function(d,e,f)').replace(/,b,/g, ',"b",')
      .replace(/,b/g, ',"b"').replace(/b,/g, '"b",').replace(/b]/g, '"b"]')
      .replace(/\[b/g, '["b"').replace(/}]/g, '"]').replace(/},/g, '}",')
      .replace(/""/g, '').replace(/length]\)}"/g, 'length])}');
  }
}

exports.default = NToken;