I'm trying to compare data from excel sheet with data obtained and extracted in matlab

I'm reading a video as text file and getting the output. Which is as follows:

""ftypisom isomiso2avc1mp41 free zmdat }e�� �!�N��+�\�~�C/x_D�0�.@�]K�M

�E

�y�5�=�� -K\b'��鉧0��gŪc� �� :�q#�e$͉���!�sv+ݓf�>~w ;"�1�P�:��l�%"

Now i want to extract the important part of that output such as ftyp, isom, free, mdat. For that i'm using strfind command to find its position and also i'll know that if the given string is present or not.

a=strfind(B,'ftyp'); //where B is my above output.

I want to create a sequence of ftyp,isom,free,mdat.... (along with its position obtained in above command )in excel sheet.

I want to compare above obtained sequence with a sequence orignally stored in excel.

eg: if orignal sequence in excel1 is ftyp,isom,free,mdat...

and the sequence obtained by extracting from the video is ftyp,isom,mdat.

I should get that my "free" is missing!

How can i do that using xmlread and xmlwrite?

Thank you in advance.

Answer 1

[Guillaume]

How can i do that using xmlread and xmlwrite?

Why do you think that xlmread or xmlwrite are relevant here? Your data is not xml. It can't be manipulated by these functions.

Why are you treating a video as text? It's clearly not text, it's binary data and should be treated as such. It would be bad luck but it's very much possible that the binary data anywhere in the file makes the string ftyp. It would be much simpler if you read the data as the structured binary data that it is (with fread calls), you wouldn't need any strfind that way.

Here are the specification of the file format, and here the MP4 extension. From that you can write a simple parser.

fid = fopen('somefile.mp4', 'r', 'b');  %MP4 always encoded in big-endian
boxes = containers.Map;  %to store all the boxes. 
while ~feof(fid)  %read until end of file
    boxsize = fread(fid, 1, 'uint32'); %get size of box
    if isempty(boxsize)
        break;  %we reached the end of file
    end
    boxtype = fread(fid, [1 4], '*char');  %strictly speaking the type is uint32, but it's readable as 4 characters
    if boxsize == 1   %extended size first
        boxsize = fread(fid, 1, 'uint64');
        boxcontent = fread(fid, [1, boxsize - 16], '*uint8');  %read content of box as uint8. We've already read 16 bytes (size, type, extended size)
    elseif boxsize == 0   %box extends to end of file
        boxcontent = fread(fid, [1, Inf], '*uint8');
    else
        boxcontent = fread(fid, [1, boxsize - 8], '*uint8'); %read content of box as uint8. We've already read 16 bytes (size, type, extended size)
    end
    %you're interested in ftyp, free, and mdat. Let's special case them
    if strcmp(boxtype, 'ftyp')
        boxcontent = struct('major_brand', char(boxcontent(1:4)), 'minor_version', char(boxcontent(5:8)), 'compatible_brands', char(boxcontent(9:end)));            
        %note: there's nothing special to do for free and mdat. free shouldn't have any content or is irrelevant. mdat content is bytes anyway
    else  
        boxcontent = struct('data', boxcontent);  %convert into structure for easy storage in map
    end
    %now store the box
    if ismember(boxtype, boxes.keys)  %already had a box of that type
        bx = boxes(boxtype); %get already stored box
        bx(end+1) = boxcontent;  %append to it
        boxes(boxtype) = bx; %and store back
    else
        boxes(boxtype) = boxcontent;  %create new box in map
    end
end
fclose(fid)

At the end of that you can easily query

boxes.keys

to see all the boxes in your file. And look at their content if they exist:

boxes('ftyp')  %guaranteed to exist
boxes('free')
boxes('mdat')

[Shrunali Paygude]

Thank you Guillaume!!

What i'm trying to do is detect forgery in videos. Whenever any video is forged, the sequence of its atoms (ex: ftyp ,mdat etc) gets changed.

So my aim is

• Extract the atoms from my video
• Compare the extracted atoms with orignal mp4 format and detect the sequence change (since im working with only mp4)

Orignal mp4 sequence: Sequence obtained from a video:(eg.)

ftyp ftyp

mdat udta

moov mdat

mvhd mvhd

udta trak......(and so on)

trak......(continued)

Either the sequence gets changed or some atoms get added/deleted.

As we can see that the sequence of atoms is disturbed when a video is forged.

In above case atom "udta" has been misplaced. This was the reason why i was trying to use excel so that i can store orignal sequence in one file.

And store the sequence extracted in another , and then compare to find the change in sequence confirming the video is forged.

There are even some mandatory and non mandatory atoms.

Detection of change in sequence is what i'm looking for.

Detection can be done as:

• Change in sequence
• Deletion of mandatory atom

Few examples of mandatory atoms are: ftyp,moov,mvhd,trak etc

If you could please help me with the code

Thank you in advance Guillaume

[Guillaume]

I'm assuming what you call atom is what the ISO standard call box. If so, I've shown you how to extract the atoms/boxes. The only thing I've not implemented is extraction of boxes within boxes.

I don't think relying on the order of the boxes is a good approach. There's no reason that an edited MP4 couldn't have the same ordering. As far as the format is concerned, I don't think the order is critical (I haven't dug into the standard to check).

In any case, it's trivial to modify the code I've written to also record the order.

fid = fopen('somefile.mp4', 'r', 'b');  %MP4 always encoded in big-endian
boxes = containers.Map;  %to store all the boxes. 
boxindex = 0;
while ~feof(fid)  %read until end of file
    boxsize = fread(fid, 1, 'uint32'); %get size of box
    if isempty(boxsize)
        break;  %we reached the end of file
    end
    boxtype = fread(fid, [1 4], '*char');  %strictly speaking the type is uint32, but it's readable as 4 characters
    if boxsize == 1   %extended size first
        boxsize = fread(fid, 1, 'uint64');
        boxcontent = fread(fid, [1, boxsize - 16], '*uint8');  %read content of box as uint8. We've already read 16 bytes (size, type, extended size)
    elseif boxsize == 0   %box extends to end of file
        boxcontent = fread(fid, [1, Inf], '*uint8');
    else
        boxcontent = fread(fid, [1, boxsize - 8], '*uint8'); %read content of box as uint8. We've already read 16 bytes (size, type, extended size)
    end
    %you're interested in ftyp, free, and mdat. Let's special case them
    if strcmp(boxtype, 'ftyp')
        boxcontent = struct('major_brand', char(boxcontent(1:4)), 'minor_version', char(boxcontent(5:8)), 'compatible_brands', char(boxcontent(9:end)));            
        %note: there's nothing special to do for free and mdat. free shouldn't have any content or is irrelevant. mdat content is bytes anyway
    else  
        boxcontent = struct('data', boxcontent);  %convert into structure for easy storage in map
    end
    %counting box order
    boxindex = boxindex + 1;
    boxcontent.index = boxindex;
    %now store the box
    if ismember(boxtype, boxes.keys)  %already had a box of that type
        bx = boxes(boxtype); %get already stored box
        bx(end+1) = boxcontent;  %append to it
        boxes(boxtype) = bx; %and store back
    else
        boxes(boxtype) = boxcontent;  %create new box in map
    end
end
fclose(fid);

After that, boxes('ftyp').index give you its position within the file, and so on.

[Shrunali Paygude]

I don't think relying on the order of the boxes is a good approach. There's no reason that an edited MP4 couldn't have the same ordering. As far as the format is concerned, I don't think the order is critical

The whole idea of detecting a forgery is based on the fact that the sequence of atoms gets changed.

We've verified that along with a paper that states it.

So the only thing that i've to do is extract and then compare with the orignal sequence.

For compairing the orignal sequence, i'll need to store it somewhere. And also check for whether my video follows that sequence(considering that there are some mandatory as well as non mandatory atoms)

[Shrunali Paygude]

Thank you Guillaume! Really apppreciate the help.

Will revert back on the code soon.

Good day!

[Guillaume]

"The whole idea of detecting a forgery is based on the fact that the sequence of atoms gets changed."

Well I can assure you that I can edit any MP4 file without changing the order of the atoms/boxes. As you can see the code I wrote can read all the boxes and store their order. I can rewrite them, with modifications, in the same order.

[Shrunali Paygude]

Thank you Guillaume!

What i was referring to is when we use softwares to edit any video, the sequence gets changed.

When editing is done without using softwares, the order of atoms/boxes can be made to be unchanged.

For now, I'm trying to do a code only for forgeries done by using softwares.

Thank you for the help!

Will surely revert back :)

Good day!

[Shrunali Paygude]

What are the changes that i'll have to do in above code if i want to read all the atoms/boxes present in the video and store them in order? (Including atoms such as free)

I displayed using keys(boxes) and values(boxes).

Now i want to get output for all atoms/boxes present in the video in sequence.

I'm unable to get all the boxes

[Guillaume]

I'm afraid I haven't got the time to implement a fully fledged parser for the mp4 format. You're going to have to dig into the data of the container boxes to get the boxes inside them. The simplest would be to delegate the parsing of the boxes to a recursive function. The parsing code would be more or less the same as I've already written.

Things to watch out for:

• Some boxes are what the ISO standard call full. Their header also include version and flag data. Then the contained boxes.
• Format of the file is big-endian. If you convert raw bytes to other types (such as uint32), don't forget to swapbytes them.
• I've not written any error handling. That's probably necessary.

The parsing is not particularly complicated. Just read the description of the boxes in the standard to know their structure. Don't be tempted to treat the raw data as text. As I said, some of the binary data (such as the video data itself) could inadvertently contain something that looks like text but isn't.

[Shrunali Paygude]

Thank you for the guidance Guillaume!

Will certainly work on the points that you've mentioned above.

Good day!