|
|
peachfuzz |
If don't model exactly, there may be too many combinations, for example, in
png.xml, if there are 20 types chuncks, there may be 20! kinds of
combinations if every chunck just appears once in every test case. 201 =
2432902008176640000. It's very large. It's combination explosion. If just
randomly give some test cases, it may miss some vulnerabilities.
I mean tree is a much better model to express the relationships than simple
liner enum. Is there some statics or papers show that tree is a good example
or is there a much better model like graph or other logic constructions to
express it? why?
I am trying to pursuade my teacher that with tree modeling relationships
between data elements is much better than simple liner model. But I can't
prove it *theoretically*. What's more, is there the relative theory in
whitebox testing?
am I boring? mike?
2009/6/28 Michael Eddington <medding...@gmail.com>
> I'm not sure what you mean by tree but it is true that information in
> one chunk could be used by another (re-png). My recommendation is
> always keep it simple. You don't need to model every exact detail,
> just enough to parse and any relations there might be.
> For PNG I typically format my pit like this:
> <DataModel name="PNG">
> <!-- Header stuff like magic -->
> <!-- Now lets parse the chunks. Don't know what order or how many
> so create an array of Choice -->
> <Choice name="Chunks" minOccurs="1" maxOccurs="10000">
> <Block name="ChunkHDR">
> <!-- Stuff for HDR -->
> </Block>
> <Block name="ChunkDAT">
> <!-- Stuff for DATA -->
> </Block>
> <!-- And so on -->
> </Choice>
> </DataModel>
> mike
> On Sun, Jun 28, 2009 at 1:17 AM, zhiyong wu<wuzhiyong0...@gmail.com>
> wrote:
> > both file types like png mp3 and networks like ftp http are expressed in
> the
> > form of tree, becase it could express paraent and brother relationships
> > between data chunks or elements.
> > Is there a best tree to express one file type like png? is it necessary
> to
> > express it out? does it exist?