Cross-Browser-Tracking-Summary-Part-4

前言

上一篇文章主要讲解了论文中具有代表性的模块的代码实现，这一篇文章主要讲论文的整体架构和处理流程，包括数据生成，数据收集，数据处理等。

整体架构

1.后端给前端下发渲染任务
2.前端进行渲染，并且将部分结果进行hash后发送给后端
3.渲染任务同时搜集浏览器信息，设备信息，一并发送给后端
4.后端接收到数据进行处理
5.生成浏览器指纹与设备指纹
6.将指纹存储到数据库，并将指纹打印到前端

模块整合/数据生成

作者使用loader.js进行模块整合，测试项目大概如下

this.testList.push(new CubeTest('normal'));
this.testList.push(new CubeTest('aa'));
this.testList.push(new CameraTest());
this.testList.push(new LineTest('normal'));
this.testList.push(new LineTest('aa'));
this.testList.push(new TextureTest(....));
this.testList.push(new TextureTest(....));
this.testList.push(new SimpleLightTest(....));
this.testList.push(new SimpleLightTest(....));
this.testList.push(new MoreLightTest(....));
this.testList.push(new TwoTexturesMoreLightTest(....));
this.testList.push(new TransparentTest(....));
this.testList.push(new LightingTest());
this.testList.push(new ClippingTest());
this.testList.push(new BubbleTest());
this.testList.push(new CompressedTextureTest());
this.testList.push(new ShadowTest());

测试结果利用dataurl传递给toServer.js进行hash处理
例如CubeTest('normal')结果如下

这十多个任务基本用来测试GPU的渲染
然后还有浏览器的字符渲染支持测试

this.asyncTests.push(new LanguageDector());

将浏览器支持的字符返回给后端

数据收集

对于数据的搜集，作者这里使用了toServer.js，大致代码解读如下
作者对如下数据进行测试和搜集

var Sender = function() {
  this.finalized = false;
  this.postData = {
    fontlist: "No Flash",
    fonts: "",
    WebGL: false,
    inc: "Undefined",
    gpu: "Undefined",
    hash: "Undefined",
    timezone: "Undefined",
    resolution: "Undefined",
    plugins: "Undefined",
    cookie: "Undefined",
    localstorage: "Undefined",
    manufacturer: "Undefined",
    gpuImgs: {},
    adBlock: "Undefined",
    cpu_cores: "Undefined", 
    canvas_test: "Undefined", 
    audio: "Undefined",
    langsDetected: [],
    video: []
  };

在调用toServer函数的时候，会传入5个参数，分别是

WebGL, inc, gpu, hash, id, dataurl

然后进行相应赋值

this.toServer = function(
      WebGL, inc, gpu, hash, id,
      dataurl) { // send messages to server and receive messages from server

    this.postData['gpuImgs'][id] = dataurl.hashCode();

    if (WebGL) {
      this.postData['WebGL'] = WebGL;
      this.postData['inc'] = inc;
      this.postData['gpu'] = gpu;
      this.postData['hash'] = hash; 
    }
  };

其中的

this.postData['inc'] = inc;
this.postData['gpu'] = gpu;

分别来自于

gl.getParameter(debugInfo.UNMASKED_VENDOR_WEBGL);
gl.getParameter(debugInfo.UNMASKED_RENDERER_WEBGL);

对于

this.postData['gpuImgs'][id] = dataurl.hashCode();
this.postData['hash'] = hash;

来自于如下运算

Uint8Array.prototype.hashCode = function() {
  var hash = 0, i, chr, len;
  if (this.length === 0)
    return hash;
  for (i = 0, len = this.length; i < len; i++) {
    chr = this[i];
    hash = ((hash << 5) - hash) + chr;
    hash |= 0; // Convert to 32bit integer
  }
  return hash;
}

这里我们知道dataurl是各种模块渲染的结果传递而来，然后利用hashcode转成数字，得到大致如下结果

紧接着

作者对4422种字体进行探测，若支持，则标记为1，不支持则标记为0
得到形如如下的数组

然后有进行了本地时间与 GMT 时间之间的时间差的搜集：

例如我的

然后作者又搜集了分辨率

这里搜集了浏览器缩放比例，浏览器的页面大小等，如下
缩放比例：80%

缩放比例：100%

浏览器页面大小：小窗和窗口最大化。缩放比例100%

然后作者又使用了navigator对象，获取了如下参数

this.postData['plugins'] = navigator.plugins;
this.postData['cookie'] = navigator.cookieEnabled;
this.postData['cpu_cores'] = navigator.hardwareConcurrency;

然后又使用了localStorage特性，判断浏览器是否支持localStorage这个属性

try {
        localStorage.setItem('test', 'test');
        localStorage.removeItem('test');
        this.postData['localstorage'] = true;
      } catch(e) {
        this.postData['localstorage'] = false;
      }

最后收集如下信息

this.postData['adBlock'] = $('#ad')[0] == null ? 'Yes' : 'No';
this.postData['canvas_test'] = Base64EncodeUrlSafe(calcSHA1(cvs_test.substring(22, cvs_test.length)));
this.postData['audio'] = audioFingerPrinting();
this.postData['langsDetected'] = get_writing_scripts();

然后将数据发送到指定ip

function startSend(postData){
        $.ajax({
          url : "http://" + ip_address + "/features",
          dataType : "json",
          contentType: 'application/json',
          type : 'POST',
          data : JSON.stringify(postData),
          success : function(data) {
            data['finished'] = true;
            parent.postMessage(data,"http://uniquemachine.org");
          },
          error: function (xhr, ajaxOptions, thrownError) {
            alert(thrownError);
          }
        });
}

数据处理

后端采用python flask框架编写

@app.route('/features', methods=['POST'])
def features():
    agent = ""
    accept = ""
    encoding = ""
    language = ""
    IP = ""

    try:
        agent = request.headers.get('User-Agent')
        accpet = request.headers.get('Accept')
        encoding = request.headers.get('Accept-Encoding')
        language = request.headers.get('Accept-Language')
        IP = request.remote_addr
    except:
        pass

但由于是demo，很多功能尚未加入，这里的跨浏览器特性就只用到了2个，难怪稳定性不是很高= =

feature_list = [
      "agent",
      "accept",
      "encoding",
      "language",
      "langsDetected",
      "resolution",
      "fonts",
      "WebGL", 
      "inc", 
      "gpu", 
      "gpuImgs", 
      "timezone", 
      "plugins", 
      "cookie", 
      "localstorage", 
      "adBlock", 
      "cpu_cores", 
      "canvas_test", 
      "audio"]

cross_feature_list = [
      "timezone",
      "fonts",
      "langsDetected",
      "audio"
      ]