Secrets Hidden in Images (Steganography) - Computerphile Computerphile English Turkish

Secrets Hidden in Images (Steganography) - ComputerphileYouTube↗ IA↗(long shot)
Education 1.1M views 22K likes 242 dislikes 04 Aug 2015 Show Contributors 3

Computerphile 2.6M subscribers N/A YouTube↗

Contributed subtitles

English GaborBartal

Click to Play Video

So cryptography is the idea of encrypting a message so that

Kriptografi bir mesajı, herkesin mesajın gönderildiğinden haberdar olmasına rağmen

although everyone knows the message has been sent they can't actually find out what it means.

kimsenin aslında mesajın içeriğini anlayamaması için şifreleme fikridir esasında.

Whereas in steganography we're trying to hide the fact that we've sent a message at all.

Oysa stenografide, mesajın gönderildiği gerçeğini de saklamaya çalışırız.

So a classic example would be if i was writing you a letter and then I wrote in

Klasik bir örnek olarak, eğer sana bir mektup yazıyor olsaydım, satırların arasına

invisible ink a whole different letter between the lines or on the other side

veya arka tarafa görünmez mürekkep ile bambaşka bir mektup yazmam

or something like that... And only you knew that that was going to be there.

veya buna benzer bir şey... Ve onun orada olduğunu yalnızca sen bileceksin.

So you get home and everyone else maybe looks at the letter

Eve gideceksin ve belki herkes mektubu okumuş olacak

and thinks "it's not very interesting at all".

ve onun önemsiz olduğunu düşünecekler.

And then of course you can uncover the secret message.

Elbette daha sonra sen, gizli mesajı açığa çıkarabilirsin.

Today we'll talk a bit about "digital image steganography"

Bugün, biraz "dijital görüntüleme steonografisi" hakkında konuşacağız

because obviously there's a huge scope for hiding things in digital images:

elbette çünkü dijital görüntüler üzerinde gizleme yapmak için geniş bir ihtiyaç var:

Images can be megabytes or more

Görüntüler birkaç megabyte veya daha fazla olabilir

and you can hide files of megabytes or more in them.

böylece içlerine birkaç megabyte veya daha büyük dosyaları saklayabiilirsin.

But of course as the the amount of steganography in images increase,

Elbette görüntü dosyalarındaki steonografi miktarı arttıkça, onu açığa

so is the attempts to try and find it.

çıkarmak için gösterilen çaba da artar.

So there's lots of statistical approaches to try to find these things as well.

Yani, bunları bulmak için tonla istatistiksel yaklaşım da mevcut.

Perhaps the most simple form of steganography an image is

Belki de bir görüntüdeki steonografinin en basit formu

"least significant bit steganography"

So if we've got a bitmap of any kind (a PNG or BMP)

Diyelim ki elimizde herhangi bir türden bitmap dosyası (PNG veya BMP) var

then we can change the lowest bits to be our message

bu durumda en alçak bitleri mesajımızı oluşturacak şekilde değiştirebiliriz.

and you'll have an almost imperceptible change

ve asıl resmin nasıl göründüğü üzerinde yaptığımız değişikliliğin

on the actual way the image looks.

fark edilmesi neredeyse imkansız olur.

It's a bit like if I change the number 800,351

Mesela 800,351 sayısını değiştirirsem

if I changed the 1 or the 51 on that

1 veya 51'i değiştirirsem bunun

it's not gonna have a massive effect... - That's exactly right, the number so big that

güçlü bir etkisi olmayacaktır... -Kesinlikle doğru, sayı öylesine büyük ki bu türden

in the grand scheme of things it makes no difference.

büyük şeylerde küçük değişiklikler fark yaratmaz.

So generally speaking we'll change (in an image) in every single byte

Genel konuşursak, bir resimdeki her bir byte'ı değiştireceğiz

we'll change the last bit or maybe the last two bits

en sondaki biti veya belki de sondaki iki biti değiştireceğiz...

if we really try to cram in a lot of data.

...eğer bir sürü veri tıkmaya çalışacak olursak.

Every byte is eight bits, we take the last two and change that to our message

Her byte sekiz bit'ten oluşur, sondaki ikisini alacağız ve mesajımızı koyacağız

in the hope no one's going to notice.

…kimsenin fark etmemesi umuduyla.

So, for every byte (that's every 8-bits) six of them are the regular image

Yani, her byte için (toplamda 8-bit) altısı görüntünün kendisine ait olacak

and two of them are our secret message,

ve diğer ikisi de bizim gizli mesajımızı oluşturacak,

so a quarter of our message is now secret.

yani, mesajımızın dörtte biri gizli

So if we have a normal pixel, it's going to be 4 bytes long

Sıradan bir pixelimiz olsaydı 4 byte uzunluğunda olurdu

(so that's one byte) so for each byte we're talking about the last two bits in that byte.

yani her byte için sondaki iki bit hakkında konuşuyoruz.

So that could be a 1, we can change it 1, change it to 0 or leave them both the same.

yani şu 1 olurdu, onu 1 yapabiliriz, 0 yapabiliriz veya her ikisini de olduğu gibi bırakabiliriz.

And what we do is we read off our message

Ve şu anda yaptığımız şey mesajımızı okumak,

so let's say our message we're trying to encrypt is 10 11 01

diyelim ki şifrelemeye çalıştığımız mesaj 10 11 01 olsun

Okay? We get to the first byte and we say: well this is great,

Tamam mı? İlk byte'a geliyoruz ve diyoruz ki: pekala bu tamam,

our first two bytes are already 1 and 0 so we don't need to change anything at all

ilk iki byte zaten 1 ve 0 bu yüzden bir şeyi değiştirmemize gerek yok

so that byte stays as it is.

yani bu bit olduğu gibi kalıyor.

So we go to the next byte, so this will be maybe red

sonra sıradaki byte'a geçiyoruz, belki kırmızı olacak

and this might be green in our pixel.

Okay? The last two bits of this byte are 0 and 1,

the two we're trying to put in from our message are 1 and 1,

so we change this one to a 1.

So by changing that second least significant bit from 0 to 1

we've just increased this value by two and we're talking about one channel

in a huge image - changing by two levels is probably not going to be too noticeable.

If we start changing the highest significant bits then that might be a problem.

All right, so I've written a program to do this

and I've tried to hide a rather large file

inside another rather large image

OK, so this is a nice picture of a tree

It's about 3 (and a bit) megapixels in size.

So this is the original image of our tree

and that is the steganographic image.

First one, into the second one.

- It's not changing! - It is changing.

When you only change the last two significant bits

of an 8 bits per channel image

you're not going to see a huge amount of detail.

if you actually do a subtraction on the images, you can see a difference

but in general it's going to be pretty imperceptible.

The really good thing to do would be to never release the original source image.

I can tell that something has changed because I've got the original and

the new steganographic image with me.

But if I just sent out an image of my dog

and I never sent out the original that the camera took

no one's going to know that it's been imperceptibly changed

because they haven't got a reference.

If you take a public domain image and change it,

it's gonna be easy to look for the original source.

- [...] - Exactly

The other thing is, it'll work better on photographs where

there's a lot of variation (in the intensity levels anyway).

So this steganographic image

has the entire works of Shakespeare buried in it

which comes to (when it's zipped up) about...

1.5 MB, something like that.

This kind of simple steganography can be detected.

This image here is an image that I've created

by taking only the last two bits of each channel.

I've scrapped all the other information. If a pixel has a value of 0, its black,

if it has a value of 3 it's white, and then it ranges in-between.

And you can see that it's a tree there, so you can see even in the first two bits that there's a tree

and the sky is particularly bland.

So if you look instead at the steganographic image

I've done the same filters to that

and you can see but the amount of noise is increased massively

because that noise is all hidden in those least 2 significant bits

So you can see if you compare the bits

from one image to the other, you can see a difference and so

hiding a messages in the least significant bits is fairly obvious

particularly if you have the original for comparison.

So this is the difference between those two images

and I've massively scaled up the difference

I mean it looks very gray.

these white pixels and black pixels are values of plus or minus 3 intensity changes.

So we're still talking very small differences over the image

and it's very evenly distributed

it's all sort of spread noisily throughout the image

- Yes so you can't actually tell that there's a tree there now.

- No, you can't tell there's a tree there. Which could be a clue!

Perhaps the more sophisticated method of hiding something in an image would be to hide it

inside the Discrete Cosine Transform Coefficients of the jpeg.

So we talked a little bit about the DCT

and how we convert an image into a series of cosine waves.

And we have coefficients saying how much of each of those waves we have.

If you change those coefficients instead of changing the raw pixel values,

you have a much less predictable effect on the image:

if you change the value of one of the alternating large current coefficients

from 202 to 201

you're gonna have a very imperceptible difference and it's going to happen over

that entire 8x8 block

so you're not going to be able to see the clear sort of steganographic noise

that we just looked at on that tree.

A common algorithm that we see in use is called JSteg.

So I see what you did there.

And what JSteg does is it goes in and if it can it will cram

DCT coefficients full of as much data as it can.

And what it does is: any coefficients aren't 0 or 1 (because they might change and be a little obvious)

so usually be low frequency ones might change up or down

and you can see again that the difference is almost imperceptible.

So here's a picture of a panda and what I've done here:

I couldn't cram in as much information as before so it's just much better than this one.

So there is the original and a steganographic image.

And I've looked at these and I've [...] a bit of difference

and you can see that again

it's very, very, very slight so these pixels again have only changed by

plus or minus 3 maybe one maybe two.

- So that's just zoomed in on the... - That's zoomed in on the difference right there

so you can see that, yes, the pictures have changed, but have not changed by a lot.

And the other crucial thing about hiding your message in the DCT coefficients:

the jpg has already completely messed up the least significant bits of the image.

So if you do an image like I did where we're looking at just the bits,

we will no longer be able to see a tree, w'll just be able to see very general jpg noise

and that will be exactly the same in our steganographic image,

so you can't do what they call a visual attack

by looking and seeing if there is a steganographic message hidden inside,

because there's no real change.

So this is the original and i'm only showing here the least two significant bits.

And you can see that they form into little blocks

[...] blocks is the 8x8 DCT blocks.

And this is the steganographic data so you can see that the blocks have changed,

but the distribution of noise throughout the image hasn't changed at all,

so it's very difficult to see there's a message buried in there

And if the message took up only a certain amount of the image

it's hard to see where in this image the message is.

You could be trying to read off every DCT coefficient

when in fact only some of them have a message in.

- If you were sending this to someone as a message...

how would they get it out?

- OK, so in general you would also encrypt the message

because, you know, you better be safe than sorry, why not use encryption.

So we encrypt our message, we put it into DCT coefficients

or in the least significant bits, and then we send it off to someone.

Now, they're going to have to have known the process that we used because if they don't,

they're gonna be looking in the wrong place, so they'll know

that we used J Stag or F5 or one of the other DCT steganography tools and

they're basically run the program they'll type in their decryption key

which will actually remove the encryption and then out comes the message.

When JSteg was invented, it was robust a visual attack

so you couldn't look at it and go: "well that clearly has been altered".

So they had to try and come up with -research had tried to come up with- some other way of

detecting that an image has had a JSteg message buried in it

and what in fact happens is that the coefficients change ever so slightly.

Because we're applying quantization to our DCT coefficients, most of them will be set to zero.

OK? And JSteg won't put anything in there, because it's too obvious;

it'll only put them in a few at the top corner that are big,

and you'll find that there's a subtle imbalance produced in where your coefficients are

so you expect most your coefficients to be 0 and then a fair few of them to be -1 or 1

and -2 and 2 to be very close to zero.

And in fact you start to get a few 3s and 4s that you weren't expecting

and the distribution of these numbers goes off a little bit and you can start to predict

that the JSteg file has been buried inside.

What's more is that this happens in each 8x8 block, so you can actually do this test

on every block and find out which blocks have messages in them, which books don't.

And you might find for example that the first 60% of the file has a message in

and then abruptly stops and that's a blatant clue

that we've got something that isn't taking up the whole image

It's has just been written sequentially into the file.

So if we take the frequency of the number of occurrences of each

DCT coefficient - so nought (0) is going to be the most common,

there may be -1 and 1, and we plot those in a graph with

frequency on the Y-axis and the DCT coefficient on the X-axis

we get what's called a histogram and that's simply a plot

of the frequency of occurrences of various things.

So you can do a histogram on an image, but you can also do a histogram on these

DCT coefficients and find out whether they've been imperceptibly changed.

Once people started routinely detecting JSteg, other people came along and

decided well that's, you know, it's too obvious, so let's try and make it be more subtle.

So what they did was they wrote DCT steganography approaches

where they pay attention to the statistics of the coefficients

and try to keep them balanced. So if you put in a 1

you try to take one out somewhere else so that you keep the histogram

and the probabilities of these coefficients occurring at the same.

And that makes it much harder to use your standard histogram analysis technique

to find out whether there's anything in the image.

but now what they can do with the power of machine learning is:

Take let's say a thousand images, 10 of which may or may not have

something buried inside and a classifier will pretty well find out which ones they are.

You just have to have a lot of positive and negative samples to throw at it.

- It all sounds wonderful but you know, [???]

- Well. yeah, so aside from spies I should say i'm not using these techniques.

you know everyone's watching....

- [???] are going through your Instagram

- Exactly, so I think one of the most common uses is digital watermarking.

So in normal steganography what we want to do is

try and hide a message as well as possible.

And then all that really matters is the person on the other side can get it

and no one else really notices.

In watermarking what we want to try and do is fingerprint to file so that we know

where it came for we know it's ours,

maybe for copyright reasons or to trace who's been distributing illegal material.

And the key to a watermark is, instead of it being as much payload as possible

so instead of trying to cram the entire works of Shakespeare into an image

what you should be doing is just a small... let's say a small logo or a small piece of text

repeated over and over so that if the image gets cropped image gets re-compressed, it's still there.

You can imagine that stock photo companies might do this

to try and make sure that people aren't distributing their files elsewhere.

And you can imagine that they would stroll through the web

looking for steganographic images embedded in their particular way.

Another case you might find if you were distributing pre-release DVDs of a film

and then it gets leaked onto the internet...

If there's steganographic data on the source buried in, you'll be able to see who it was that leaked it.

- Each file could be tailored... - Each file could be tailored with

the person they originally sent that file to

and then when that particular find their way onto the Internet, that person is going to be in trouble.

What was vital in recreating this image is now gone and we're not going to get it back

and in fact that's exactly what you do see so if we show the actual output here

we can see [...] is kind of visible but it's been completely dwarfed by

all this random noise that's been added...

Similar videos:

Getting Sorted & Big O Notation - Computerphile
Computerphile
195.5K views 4.6K likes 41 dislikes 18 Jun 201356% match

Mac or PC? - Computerphile
Computerphile
587.9K views 9.7K likes 314 dislikes 21 May 201435% match

TOR Hidden Services - Computerphile
Computerphile
859.3K views 16.2K likes 242 dislikes 09 Jun 201733% match

What is Bootstrapping? - Computerphile
Computerphile
375.1K views 5.5K likes 287 dislikes 19 Mar 201433% match

Resizing Images - Computerphile
Computerphile
360.9K views 8.5K likes 115 dislikes 02 Sep 201631% match

What is a Monad? - Computerphile
Computerphile
357.7K views 7.6K likes 1.2K dislikes 24 Nov 201731% match

PDF, What is it FOR? - Computerphile
Computerphile
409K views 12K likes 144 dislikes 16 Sep 201630% match

Why C is so Influential - Computerphile
Computerphile
1.6M views 22.5K likes 1.6K dislikes 25 Aug 201730% match

Where HTML beats C? - Computerphile
Computerphile
551.2K views 11K likes 1.3K dislikes 27 May 201629% match

Lambda Calculus - Computerphile
Computerphile
803.9K views 19.5K likes 567 dislikes 27 Jan 201729% match

Public Key Cryptography - Computerphile
Computerphile
669.9K views 18K likes 224 dislikes 22 Jul 201429% match

Computing Limit - Computerphile
Computerphile
358.4K views 11.3K likes 443 dislikes 03 Nov 201728% match

Where GREP Came From - Computerphile
Computerphile
644.1K views 23.4K likes 152 dislikes 06 Jul 201828% match

Fuzzy Logic - Computerphile
Computerphile
335.5K views 5.5K likes 185 dislikes 11 Jun 201428% match

Program Correctness - Computerphile
Computerphile
48.8K views 2K likes 253 dislikes 27 Nov 202028% match

Internationalis(z)ing Code - Computerphile
Computerphile
911.6K views 32.5K likes 495 dislikes 04 Aug 201427% match

Password Cracking - Computerphile
Computerphile
3M views 60.3K likes 1K dislikes 13 Jul 201627% match

Parsing Explained - Computerphile
Computerphile
134.2K views 3.6K likes 45 dislikes 13 Nov 201927% match

P vs NP on TV - Computerphile
Computerphile
527.4K views 7.8K likes 537 dislikes 08 Feb 201427% match

Deep Learning - Computerphile
Computerphile
392.2K views 6.1K likes 280 dislikes 19 Apr 201627% match

What on Earth is Recursion? - Computerphile
Computerphile
565.8K views 16.8K likes 123 dislikes 16 May 201427% match

ALGOL 60 at 60 - Computerphile
Computerphile
73.7K views 3.1K likes 20 dislikes 04 Jun 202027% match

How to Choose a Password - Computerphile
Computerphile
1M views 21.1K likes 235 dislikes 20 Jul 201626% match

Unicorn AI - Computerphile
Computerphile
339.5K views 7.8K likes 189 dislikes 04 Jul 201926% match

Entropy in Compression - Computerphile
Computerphile
332.2K views 7.1K likes 69 dislikes 02 Jul 201326% match

Reflections on Trusting Trust - Computerphile
Computerphile
66K views 3.7K likes -1 dislikes 23 Mar 202126% match

Slow Loris Attack - Computerphile
Computerphile
962.6K views 31.1K likes 265 dislikes 09 Nov 201626% match

Dijkstra's Algorithm - Computerphile
Computerphile
1.1M views 23.1K likes 392 dislikes 04 Jan 201724% match

Onion Routing - Computerphile
Computerphile
1.1M views 24.4K likes 255 dislikes 31 May 201724% match

Maze Solving - Computerphile
Computerphile
1M views 21.8K likes 295 dislikes 24 Feb 201724% match

Settings Privacy About Contact Donate