My name is Yoshinori Takesako. It's very long name. so call me

My name is Yoshinori Takesako. It's very long

name. so call me "Yoshi". I came from Tokyo,

Japan.

I am a chairman of the SECCON. SECCON is a

largest security CTF contest in Japan. I am an

CTF organizer and a challenge creator. I am

also on the Open Web Application Security

Project - OWASP Japan advisory board.

And I am the review board for the CODE BLUE

which is a biggest international security

conference in Japan.

In this year, About 2,500 people took part in the

SECCON CTF qualifier from 58 countries around

the world. We held the international SECCON

CTF Final competition in this year at Tokyo,

Japan. Finally, Korean hacker team had won.

That was great.

I want to show the NIRVANA-KAI SECCON

Customized Mk-II. This is a real time

visualization system for attack and defense

battle of CTF. This real time visualization system

was developed by National Institute of

Information and Communications Technology -

NICT in Japan.

Okay, I talk about “Backdooring MS Office

documents with secret master keys”. We made

a lot of CTF challenges such as a XSS, reversing,

pwn, cryptography at SECCON CTF project.

when I created some cryptography challenges,

I found this backdoor problem. Microsoft Office

twenty ten or later version employ "Agile

Encryption" algorithm in their OOX documents.

We found a vulnerability in the file format

specification that can allow an attacker to later

decrypt strongly encrypted documents without

the password. This is possible by tricking MS

Office into creating an undetectable secret

master key when it creates encrypted

documents.

Microsoft has standardized the OOX file format

by ECMA international. It is not "Open Office"

XML format, Open Office is a rival application

for Microsoft. So “Office Open" XML format is

correct.

You can see at the DOCX suffix from filename.

And you know it is just a zip archive file.

However, when you encrypt a DOCX file, it will

become an old classical DOC FILE format. This

encrypted.DOCX file has a DOCX suffix. But, it is

not the zip archive file. You can see at file hex

dump header, "D0 CF 11 E0“. it is DOCFILE's leet

character!

Old classical DOC FILE format has been

standardized as the MS-CFB. This specifications

documentation was opened by Microsoft. I

think it is a great job.

MS-CFB file format has some mini FAT sectors.

Mini FAT has a 64-byte small sector size. mini

FAT sectors are in a standard chain in the FAT.

This is a figure of File layout of encrypted DOCX

file. Any encrypted DOCX file have a these file

in mini FAT. EncryptionPackage is an binary file

which was encrypted from original DOCX zip

file object. EncryptionInfo is very important

information for these encryption parameters.

Microsoft opened this Office cryptography

Cryptography Structure as [MS-OFFCRYPTO]. We

tried to read the MS-OFFCRYPTO document

carefully.

Yes, we can read the binary of DOC FILE!

If you want to protect your document with

passwords. In Microsoft Office, you can use

passwords to help prevent other people from

opening or modifying your documents.

Select the Protect Document menu on Info tab.

You can choose some submenu.

When you select "Encrypt with Password", the

Encrypt Document dialog box appears.

In the Password box, type a password. And

confirm password. Then Encrypted.docx is

saved.

There are another manipulation.

You can select SaveAs menu.

And push the Tool button and select

GeneralOption.

Then you can input the password just the same

way.

It's important to know that you don't forget the

password. Because Microsoft cannot retrieve

your forgotten passwords. If you forget the

password, we can not retrieve the original

documents.

But, is it true? I cannot decrypt actually?

For when this occurs , there are password

recovery software. oclHashcat is one of the

famous password recovery tools by command

line.

If you want to crack password protected MS

Office documents, type this command.

Recently oclHashcat supports GP-GPU power,

and supported new Office document OOX file

format.

Before you have to extract hash from encrypted

file by office2john.py.

There are another password recovery software

with simple User interface. I used the

Passcovery commercial edition which is very

powerful password recovery tools.

It's very simple graphic user interface. Only

clicking.

I evaluated comparing the decryption time of

password cracking. There are some encryption

file format. Classic Zip and AES Zip nad old

DOCFILE and new DOCX files.

DOCX files are very strong against Brute-force

attack.

Password consists of Latin small characters an

Latin capital characters and digits and special

symbols characters. If the password length is 8.

Time required to decrypt the encrypted Classic

ZIP file by brute force attack was 15 minutes. Its

encryption key bit is only 96.

WinZIP have a long AES encryption key, Time

required to decrypt the encrypted new WinZIP

file by brute force attack was 6 days. Time

required of brute force attacking has increased

gradually with Office version is newer.

Password consists of Latin small characters an

Latin capital characters and digits characters. If

the password length is 8. DOCX’s time required

of brute force attacking was about Twenty

thousand years with Office version is twenty

thirteen.

Password consists of 93 letters which include

Latin small characters and Latin capital

characters and digits and special symbols

characters. If the password length is 8. Office

twenty thirteen DOCX’s time required of brute

force attacking was about Sixty-seven million

years. If the password length is 10, you will not

be able to decrypt even coming the next Big

Bang.

This program code is password checking and

decoding algorithm. Please attention the line

that the secretkey is used. decData is

dependent on only secretkey and keydatasalt.

It is not dependent on password.

This is a figure of dependency of values in

decoding.

Decoded contents is dependent on only

secretkey and keydata.saltValuy. It is not

dependent on password. I think that it is a

problem.

This is a figure of dependency of values in

encoding.

There are problem with generating the

secretKey.

The secretKey used in AES encryption needs to

create an unique key with random data.

If the key is long enough and was created with

truly random data then it is thought to be

extremely difficult to crack.

However, if the secretKey was chosen in a

predictable manner then it will be easy to

crack.

The integrity of secure random generators (both

software and hardware based) are imperative

for strong encryption.

I would like to introduce my friend. Shigeo

Mitsunari is a software developer and

researcher at Cybozu Labs company. He

developed this msoffice-crypt.exe tools.

I was working together with him. He is a co-

author of this paper.

We made the encryption and decryption tools

for new Microsoft Office DOCX and XLSX and

PPTX files.

msoffice-crypt command has a decode with

inputed password options. It is –d and –p.

We made a decode with master key options. Is

is –k.

And we made a encode option. It is –e. Then

we have –e and –k and –p options. We can

make two encrypted files by another password

with same master key. It is a backdoor.

1. In this demo, demo1.xlsx is encrypted with

the password "pass". The target software is

MS Excel twenty-thirteen.

2. demo2.xlsx is encrypted with another

password "pass1234".

3. However, MS Office was manipulated to

implant a hidden master key when these

files were created.

4. Therefore, these files can be easily

decrypted by the same master key

without any need to brute-force the

password.

5. In this example, the master key is set to

"001122...FF0011...FF".

IT admin can "unlock" the password-protected

OOXML Word, Excel and PowerPoint files for a

user and then either leave the file without

password protection! (it is official)

attack vector is that some attacker can

replace the random generator function by

Win32 API hooking.

There are so many API hooking techniques. IAT

Import Address Table function hooking is one of

the famous Windows API hooking techniques.

And more over, there are WinAPIOverride thirty-

two and sixty-four application. It’s very nice

software. I like it.

Microsoft Research created general purpose

function hooking library “Detours”.

It can easily hook the application by DLL

injection.

In this case, I can hook the CryptGenRandom

function on Advapi32.dll. Then hooked

CryptGenRandom function always return the

fixed value. 0x33

In other case, I can hook the CPGenRandom

function on old Windows API. Hooked

CPGenRandom function always return the fixed

value which is not random value.

In another case, I can hook the

rtl_random_getBytes function on sal3.dll which

is used by LibreOffice application. I can control

the randomness on my own computer.

attack vector is that some attackers can

replace the random generator in embedded

hardware chips.

Intel developed the RdRand instruction in the

hardware chip. Core i7 so on. It generate truly

random by Intel's new hardware chips.

The pseudo-device /dev/random generates a

virtually endless stream of random numbers on

GNU/Linux systems. RdRand is an instruction

found in modern Intel CPU chips that stashes a

"high-quality and high-performance entropy"

generated random number in a given CPU

are vital in producing secure session keys, new

public-private keys and padding in modern

encryption technology.

If some government intelligence agencies have

managed to persuade Intel to hobble that

instruction. The strength of encryption

algorithms will be weak on that random data.

Linus Torvalds's answer is very simple.

we use rdrand as one of many inputs into the

random pool, and we use it as a way to

improve that random pool. So even if rdrand

were to be back-doored by some government

intelligence agencies, our use of rdrand

actually improves the quality of the random

numbers you get from /dev/random.

We can get the source code of Linux. And This

is because it can be verified binaries on your

own machines. It’s very important that Linux is a

open source software.

However, what is in the cloud environments?

attack vector is that some attackers can use

the predictable number generator secretly in

cloud environments.

Recently, Microsoft released an Office online.

You can try this one as Office twenty-sixteen

preview edition. The Office application will be

on the Microsoft's cloud system. I think that we

can not stop these cloud system movements

now. We should check how the cloud

encryption algorithm and encryption system is

safety. Some industry companies become to

have an interest in safety encryption system. I

think that it is important things. Linux is a open

source, but Microsoft product is closed source.

Recent MS Office twenty-ten or later version’s

documents are normally encrypted very

strongly, making them difficult to brute force

attacks. However, there are techniques some

attacker can use to secretly backdoor these

encrypted documents to make them trivial to

decrypt. Cloud environments may be more

dangerous than thought as it is not possible for

users to confirm the security of their encryption.

And it would be easy for cloud providers to

backdoor encryption in undetectable ways.

If advanced attackers can access to those

cloud providers, it will become a serious

problem.

Thank you for your attention. And I want to say

thanks for some supported members. That’s all.