Click here to Skip to main content
Click here to Skip to main content

HTTP Monitor for Webbrowser Control

, 27 Sep 2011 CPOL
Rate this:
Please Sign up or sign in to vote.
The ATL COM DLL that captures requests from individual Webbrowser Control
Demo

Introduction 

There are various tools available to monitor HTTP traffic that is being sent and received from different processes. Fiddler is one such good example. All these programs open a port and filter HTTP traffic based on process id. But if a C# app consists of multiple browsers, they fail to identify which request was sent by which browser.

The C# browser control only provides Navigating and Navigated events and does not give any idea about the requests that it sends (e.g. loading of images, etc.).

This article provides an ATL COM DLL that can monitor HTTP traffic from individual browsers.  

Background

While working on a project which required the same, I stumbled upon PassThruApp by Igor Tandetnik.

csExWBDLMan.dll ("csExWBDLMan COM library" from The most complete C# Webbrowser wrapper control) is one implementation of PassThru App that provides the requests, but does not provide detailed information of redirections, data received cookies, etc. in the requests. So, I decided to write a custom code just for monitoring HTTP traffic based on both PassThru App and csExWBDLMan.dll.

About PassThru App

"It is an object that implements both sides of URL moniker-to-APP communication, that is, it implements both IInternetProtocol and IInternetProtocolSink / IInternetBindInfo. We register it as a temporary handler for a standard protocol, such as HTTP. Now whenever an HTTP request needs to be sent, URL moniker will create an instance of our pAPP and ask it to do the job. The pAPP then creates an instance of a standard APP for the protocol in question (I call it a target APP, or tAPP...) and acts as its client. At this point, our pAPP becomes a proverbial man-in-the-middle. In the simplest case, any method call made by URL Moniker on pAPP is forwarded to tAPP, and any method call made by tAPP on pAPP is forwarded back to URL Moniker. The pAPP gets to observe, and if desired modify, every bit of information relevant to this request passing back and forth between the moniker and the tAPP. QED" - Igor Tandetnik

The Code

The code extends classes provided by PassThru App. There are two main classes:

  1. MonitorSink - MonitorSink extends PassthroughAPP::CInternetProtocolSinkWithSP that implements IInternetProtocolSink
  2. CTestAPP - CTestAPP extends PassthroughAPP::CInternetProtocol that implements IInternetProtocol
class MonitorSink :
public PassthroughAPP::CInternetProtocolSinkWithSP<MonitorSink>,
    public IHttpNegotiate
{.. 
class CTestAPP :
	public PassthroughAPP::CInternetProtocol<TestStartPolicy>
{..

Now we can intercept requests using:

  1. Request - MonitorSink::BeginningTransaction
  2. Response - MonitorSink::OnResponse
  3. Redirection, Cookies Transferred, Error, Cache Loaded, etc.- MonitorSink::ReportProgress when ulStatusCode from IInternetProtocolSink->ReportProgress(ulStatusCode.. is BINDSTATUS_REDIRECTING etc. depending on the information required.
  4. Data received - CTestAPP::Read

But the problem is that we are using Asynchronous Pluggable Protocol and all the requests are done asynchronously. So we get all information, but cannot say which response belonged to which request. Moreover, the data is received asynchronously in chunks.

The best solution is that if we get unique id for a transaction (i.e. unique id attached request, response and data received), then we will be able to weave the async calls back together. Here we get lucky.

  1. IInternetBindInfo for a request is most of the times unique and is available in all the methods. But sometimes, it is reused by the Browser.
  2. IHTMLWindow2 exists and is unique in case request is sent from iframe.
  3. The Url to which request is made is also most likely to be unique.

So if we create an id from all of them, we will get a unique id for a transaction. Now we do not limit to this. If there are multiple iframes present on the page, then we can traverse the page and tell which iframe sent which request based on referrer. In addition, each iframe fires BeforeNavigate and NavigationComplete events on navigating. Objects of the iframes are available from InternetExplorer interface. If you link all the available information, you can draw a complete picture of:

  • What is the hierarchy of the iframes on the page
  • Which iframe navigated to which URLs
  • What were the requests sent by a particular iframe
  • What URLs failed to load
  • Which URLs used files from the local computer (with their actual file location)
  • What headers, cookies were sent and received
  • How much time did each request take, etc.

Limitations

  1. If you set Silent=true in native COM control or ScriptErrorsSuppressed=true in .NET WebBrowser, the above code stops working.
  2. The assumption behind filtering requests based on iframes is that every iframe on a page has a different location.
  3. One more assumption is that a flash from a particular URL is loaded only in one webrowser control (if many controls are present) and only in one page. Otherwise the requests sent from flash control will get mixed up. The reason is that requests sent by flash show referer as flash object instead of the page URL and we cannot determine which flash (from which control) actually sent the request.

Using the Code

When you attach your browser with HttpMonitor.dll, on each request, response, etc. an event is fired with all the required arguments. There are twelve plus one events available:

  • OnRequest(int id, int containerId, string url, 
    	string headers, string method, object postData)
  • OnRedirect(int id, int containerId, int redirectedId, string url, 
    	string redirectedUrl, string responseHeaders, string requestHeaders)
  • OnResponse(int id, int containerId, string url, int responseCode, string headers)
  • OnDataRecieved(int id, int containerId, string url, object data, bool isComplete)
  • OnCookieSent(int id, int containerId, string url, string cookies)
  • OnCookieRecieved(int id, int containerId, string url, string cookies)
  • OnMimeTypeAvailable(int id, int containerId, string url, string type)
  • OnCacheLoaded(int id, int containerId, string url, string location)
  • OnP3PHeaderRecieved(int id, int containerId, string url, string p3PHeader)
  • OnError(int id, int containerId, string url, int result, int errorCode)
  • OnProgress(int id, int containerId, string url, 
    	int grfBSCF, uint progress, uint progressMax)
  • GetIServiceProviderOnStart(int id, int containerId, string url, int ptr)

One more event is available:

  • ConfirmRequest(int id, int containerId, string url, 
    	int totalInstances, ref bool itsMine)

If request is of flash object or sent from flash object, then the DLL is not able to determine which browser sent it. It fires the above event and asks you to look into your request logs and suggest whether it belongs to your browser or not based on container id. Sample implementation is provided in demo app. If you set itsMine=true by default, you can trace all the requests made by the current process.

One common mistake is to think CHttpMon is used only for one transaction at a time. If CHttpMon contains a private variable, it will be shared in diff requests and we cannot store data for any one request in it. Also referer for the requests made by flash object is location of the flash object instead of the page. So we need to keep track of these flash objects. Anyways all these objects behave just like Microsoft intended and not like how we want them to be. Most of this is undocumented and we need to do A/B test to find out how they actually work.

IHTMLWindow2 is available for requests made by iframes. This can also be exploited further (i.e. events like new OnPageRequest). I was hoping that I will receive this in all requests and create containerId using IHtmlWindow2 and referer that will be totally unique but it looks like that is not possible :(

You will first need to navigate to about:blank so that "Internet Explorer_Server" window is available. Then, attach the browser using handle of "Internet Explorer_Server window".

if (monitor == null)
{
     monitor = new HttpMonitorLib.HttpMonClass();
     monitor.IEWindow = GetTopWindow(GetTopWindow
		(GetTopWindow(webBrowser1.Handle))).ToInt32();
     monitor.OnRequest += new HttpMonitorLib._IHttpMonEvents_OnRequestEventHandler
			(monitor_OnRequest);
.
.

For example, the following function will be executed whenever the browser sends a request.

private void monitor_OnRequest(int id, int containerId, string url,
	string headers, string method, object postData)
{
//code here
}   

The id specifies unique id associated with that particular HTTP transaction and containerId is the uniqueId of the iframe sending the request. Basically, it is hash of the iframe's current location.

What's New

  1. All request headers were not reported in OnRequest event in the last version as all the headers are only available after BINDSTATUS_SENDINGREQUEST.
  2. Added OnProgress and GetIServiceProviderOnStart events.
  3. Modified demo app implementation so that request of an iframe belongs to its page instance instead of its parent's page instance.
  4. Code refactoring/optimization in both demo app and httpmonitor DLL.

Demo App

The demo app provides the implementation of HttpMonitor to detect complete page navigations. The main class page accepts "Internet Explorer_Server" as parameter. It can be considered wrapper of InternetExplorer interface. The class has the following properties:

  • Children - List of children pages
  • Entries - List of requests/response sent by this page/iframe
  • Navigations - Navigations done by this page or iframe
  • AllIEWithNavigations - All instances of InternetExplorer interfaces with their navigations
  • Webrowser - Actual instance of InternetExplorer interface of current page
  • AllEntries - All requests/response sent by current control
  • AllPages - Instances of all Page objects

The reason I put containerId as integer instead of string (as containerId is hash of referral URL) is that I thought that somehow I will be able to get IHTMLWindow2 pointer and I will pass this as containerId. So far, I have been unsuccessful. If anybody is able to figure out how to do this, please do tell me.

There are crude ways to achieve this, for e.g., in BeforeNavigate2 event, set Cancel=true and renavigate to
url + "IHTMLWindow2=<pointer to IHTMLWindow2>" in query string
or navigate with "IHTMLWindow2:<pointer to IHTMLWindow2>" in headers
and parse requests to get the value. But firstly, this breaks the regular navigation and secondly flash object's requests again do not persist these values.

I want to thank Igor Tandetnik for his wonderful PassThruApp that made all this possible.

License

This article, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Share

About the Author

abhinav_soni

India India
No Biography provided

Comments and Discussions

 
You must Sign In to use this message board.
    Spacing  Noise  Layout  Per page   
GeneralMy vote of 1 PinmemberCary Ng7-Jan-14 22:25 
Questionhttps urls PinmemberJerichoSteele13-Dec-13 6:44 
GeneralMy vote of 1 Pinmemberynw0302811-Nov-13 6:15 
GeneralMy vote of 1 Pinmemberynw0302811-Nov-13 6:14 
QuestionRaise System.AccessViolationException & how to stop monitor Pinmemberoliverbuaa23-Sep-13 21:10 
QuestionHow to debug this dll ,when I try to debug,the breakpoint not valid Pinmemberframework327-Aug-13 4:35 
QuestionHTTP Monitor not working in live IE page PinmemberGaurangPatel12-Mar-13 22:07 
AnswerRe: HTTP Monitor not working in live IE page Pinmembervivek_acharya21@yahoo.com12-Mar-13 23:33 
GeneralRe: HTTP Monitor not working in live IE page Pinmemberabhinav_soni2-Apr-13 20:55 
QuestionHttpMonitorLib is missing in the Demo.Zip PinmemberSpark2Fly28-Jan-13 9:31 
QuestionCannot add to Visual Studio 2010 PinmemberNong Tien Bac16-Aug-12 22:51 
QuestionPluggable Namespace VS pluggable protocol PinmemberMember 128978712-Dec-11 15:09 
QuestionHttp monitor and BHO Pinmemberkstxx29-Nov-11 12:20 
AnswerRe: Http monitor and BHO Pinmemberluismc9-Dec-11 9:50 
GeneralRe: Http monitor and BHO Pinmemberabhinav_soni10-Dec-11 6:43 
GeneralRe: Http monitor and BHO [modified] Pinmemberluismc12-Dec-11 0:40 
GeneralRe: Http monitor and BHO Pinmemberabhinav_soni18-Dec-11 23:43 
GeneralRe: Http monitor and BHO Pinmemberluismc20-Dec-11 7:44 
GeneralRe: Http monitor and BHO Pinmemberabhinav_soni21-Dec-11 23:22 
GeneralRe: Http monitor and BHO Pinmembervivek_acharya21@yahoo.com12-Mar-13 22:20 
BugCrashes on some URLs PinmemberOceansp26-Oct-11 1:03 
QuestionCertificate messages [modified] Pinmembermarc ochsenmeier24-Oct-11 22:45 
Suggestionfor each request ,there is a CTestAPP,uhm? Pinmemberxinkmt22-Sep-11 18:56 
GeneralRe: for each request ,there is a CTestAPP,uhm? Pinmemberabhinav_soni24-Sep-11 6:48 
BugPassthruAPP.zip has two "assert fail"s [modified] Pinmemberxinkmt22-Sep-11 16:28 

General General    News News    Suggestion Suggestion    Question Question    Bug Bug    Answer Answer    Joke Joke    Rant Rant    Admin Admin   

Use Ctrl+Left/Right to switch messages, Ctrl+Up/Down to switch threads, Ctrl+Shift+Left/Right to switch pages.

| Advertise | Privacy | Terms of Use | Mobile
Web03 | 2.8.150123.1 | Last Updated 27 Sep 2011
Article Copyright 2011 by abhinav_soni
Everything else Copyright © CodeProject, 1999-2015
Layout: fixed | fluid