SeaUrchinBot / google-security-research

Automatically exported from code.google.com/p/google-security-research
1 stars 0 forks source link

Google Chrome: Privilege Escalation from Renderer Process to Browser Process #664

Open GoogleCodeExporter opened 8 years ago

GoogleCodeExporter commented 8 years ago
There is an overflow in the ui::PlatformCursor WebCursor::GetPlatformCursor 
method. In 
src/content/common/cursors/webcursor_aurax11.cc&q=webcursor_aurax11.cc, there 
is the following code:

bitmap.allocN32Pixels(custom_size_.width(), custom_size_.height());
memcpy(bitmap.getAddr32(0, 0), custom_data_.data(), custom_data_.size());

The bitmap buffer is allocated based on the width and height of the 
custom_size_, but the memcpy is performed using the size of the custom_data_.

These values are set during WebCursor deserialization in 
src/content/common/cursors/webcursor.cc in WebCursor::Deserialize.

custom_size_ is set from two integers that a deserialized from a message and 
can be between 0 and 1024. custom_data_ is set from a vector that is 
deserialized, and can be any size, unrelated to the width and height. The 
custom_data_ is verified not to be smaller than the expected pixel buffer based 
on the width and height, but can be longer.

GetPlatformCursor is called indirectly by RenderWidgetHostImpl::OnSetCursor, 
which is called in response to a  ViewHostMsg_SetCursor message from the 
renderer.

The issue above is in the x11 implementation, but it appears also affect other 
platform-specific implementations other than the Windows one, which instead 
reads out of bounds.

I recommend this issue be fixed by changing the check in WebCursor::Deserialize:

if (size_x * size_y * 4 > data_len)
    return false;

to

if (size_x * size_y * 4 != data_len)
    return false;

to prevent the issue in all platform-specific implementations.

To reproduce the issue replace WebCursor::Serialize with:

bool WebCursor::Serialize(base::Pickle* pickle) const {

  if(type_ == WebCursorInfo::TypeCustom){
  LOG(WARNING) << "IN SERIALIZE\n";
  if (!pickle->WriteInt(type_) ||
      !pickle->WriteInt(hotspot_.x()) ||
      !pickle->WriteInt(hotspot_.y()) ||
      !pickle->WriteInt(2) ||
      !pickle->WriteInt(1) ||
      !pickle->WriteFloat(custom_scale_))
     return false;
   }else{

     if (!pickle->WriteInt(type_) ||
      !pickle->WriteInt(hotspot_.x()) ||
      !pickle->WriteInt(hotspot_.y()) ||
      !pickle->WriteInt(custom_size_.width()) ||
      !pickle->WriteInt(custom_size_.height()) ||
      !pickle->WriteFloat(custom_scale_))
    return false;

  }
  const char* data = NULL;
  if (!custom_data_.empty())
    data = &custom_data_[0];
  if (!pickle->WriteData(data, custom_data_.size()))
    return false;

  return SerializePlatformData(pickle);
}

and visit the attached html page, with the attached image in the same directory.

This bug is subject to a 90 day disclosure deadline. If 90 days elapse
without a broadly available patch, then the bug report will automatically
become visible to the public.

Original issue reported on code.google.com by natashe...@google.com on 3 Dec 2015 at 1:48

Attachments:

GoogleCodeExporter commented 8 years ago
This is Chrome issue https://code.google.com/p/chromium/issues/detail?id=565023

Original comment by natashe...@google.com on 3 Dec 2015 at 1:50

GoogleCodeExporter commented 8 years ago

Original comment by natashe...@google.com on 17 Dec 2015 at 11:18

GoogleCodeExporter commented 8 years ago
Why is this not restricted? The linked Chromium issue is currently restricted...

Original comment by phistuck on 30 Dec 2015 at 7:17

GoogleCodeExporter commented 8 years ago
The fix for this was released with Chrome 47.0.2526.106

Original comment by wfh@chromium.org on 30 Dec 2015 at 9:00

GoogleCodeExporter commented 8 years ago
If this is not restricted, the Chromium issue should not be restricted (barring 
sensitive information within the issue, of course)...

Original comment by phistuck on 30 Dec 2015 at 9:37

GoogleCodeExporter commented 8 years ago
Nothing new or strange with different projects having different disclosure 
practices.

P0 choose to disclose immediately the issue is fixed, but Chromium opens up 
bugs automatically a number of weeks after the issue is fixed in Stable.

Original comment by wfh@chromium.org on 30 Dec 2015 at 12:48