Skippipe: Skipping the watermark in digital content

Gerard Fuguet ([email protected])

Abstract
Words written on a physical or digital format, has an origin. The source can come from
an individual or group of these who wants to transmit, document something to show it.
The owners are free to impose their rules, like the protection of their works, generating a
copyright or/and a watermark subject to a subordinate. It sometimes can create extra
content; words, images or growing a file (if it’s a digital document). The processes that
we focus on this white paper is to remove this extra content from a digital document, thus
avoiding the obstacles that degrade it, which makes interpretation difficult.
We will demonstrate how to remove easily the student watermark from a Microsoft MOC
(official courses) that uses Skillpipe as platform to read their digital courses. This situation
motivated me to write it.
The intention is that owners of their works, takes consciousness protecting well their
content but without disturbing it with possible distortions once protected.

1
 Table of Contents

1. Motivation ................................................................................................. 3
2. Skillpipe: The eReader .............................................................................. 5
2.1. Web App: The Past ............................................................................. 6
2.1.1. The Cook ....................................................................................... 7
2.2. Web App: Present ............................................................................... 8
2.2.1. The Kitchen Assistant ................................................................... 9
2.3. Native Application ............................................................................ 10
2.3.1. Dusting It .................................................................................... 10
2.3.2. Caching It .................................................................................... 12
2.3.3. BDB2 .......................................................................................... 16
2.3.4. Tracking It ................................................................................... 18
3. Water-Engine-Mark ................................................................................ 24
3.1. Pipe Factory ...................................................................................... 24
3.2. Types ................................................................................................. 25
4. Pipe Leak ................................................................................................. 26
4.1. Manual Repair ................................................................................... 26
4.1.1. Cast Iron (Old) ............................................................................ 26
4.1.2. PVC (New).................................................................................. 29
4.2. Semi-Auto Repair ............................................................................. 31
4.2.1. Ad Blocking ................................................................................ 31
5. Conclusions ............................................................................................. 33
6. References ............................................................................................... 38

2
1. Motivation
Since “minus 0 ages”, human cannot stop learning. If you think that, in some type of
thing cannot be learned more about it, there is something you need to still see, although
your eyes cannot perceive it.
We wanted to know, we wanted to learn more about some Microsoft products. You can
achieve this lot of beautiful information on Internet as Microsoft is popular in IT world,
but if you look for a some special (be a specialist) you need to join to the rules of academic
world. The time to study the official courses started, they are called: MOC [1].
The chosen Microsoft partner offers many facilities to study. I chose the face-to-face
option in an official/authorized center. Despite having an on-site teacher, the rest of the
material was online (books and labs) and the documentation was subject to being read
with an application; you can only do it with the help of Skillpipe [2].
My big problem originated under my offline study needs. The platform offers the print
functionality of the material, but some problems bothered me:
- Watermark acts as an image on every page and this generates extra weight in the
document (more than 1000% most of the times).

Figure 1: At left, the properties of document with watermark; at right, without watermark

- When you select text around the watermark, part of it can’t be catch.

3
 Figure 2: Comparison selecting all the text with and without watermark respectively

- Vertical bars in the corners are generated and produce distortions if some
words/letters are in same “road” of the bars.

4
 Figure 3: Distortions observed at corners with and without watermark respectively

I think I had enough arguments to delete the watermark from my documents, the intention
was never to distribute illegally, of course! :P I wanted to learn comfortably without
forcing my eyes, select the text for copy & paste as normal…

I threw some questions to the trainer:

- Is there any solution to select all the text under PDF format?
- No
- Well, and… any trick to delete the watermark?
- No, all you can do, is print by modules.

OK, seems nothing to do with it, but didn’t convince me. I'm not against the watermarks
for protect the rights, but I don't like it to end up as an annoying “leak”. Somehow, I
needed to “repair” the situation, cleaning or “drying” the watermark with whatever.

2. Skillpipe: The eReader

Is the software, the tool for read the books/documentation via web-based browser,
offering only online mode [3]. This isn’t developed by Microsoft, is made by Arvato
Supply Chain Solutions – Arvato Distribution GmbH [4]. Delegate the reader part to a
third party is not bad, if Microsoft take decision to not to develop the reader internally

5
should be for a variety of reasons that are out of scope here, but externalize some of the
things is maybe better if you have a lack of experience (is impossible knowing all when
the live is limited!).
The use of the web tool is very easy, maybe this could be one of the reasons to be the
perfect candidate for all people. They also have a good user guide explaining all
functionalities and the one we are interested on, is in how the web app handle the printing
option. We will see two types/manners of behaviors; one is an old methodology used in
a previous version (I don’t remember the code number), other is in nowadays or version:
3.0.
Also, we will take a short walk for the on-premises app that it was intended to be used for
offline reading.

2.1. Web App: The Past

This was tested, for about, on the begins of the 2017 year.
The idea is to print all in one “shot”, right? So, after select one assigned course, identify
the print button is easy, is located at the left-bottom of the page as shown in figure 4
below:

Figure 4: Printing functionality (old version)

6
Two popups appear: New tab in browser and then the print assistant.

Figure 5: Printing assistant through new tab

Remember what the trainer said in chapter 1, the book cannot be printed entirely in one
time, only the current module prints [5]. You need to choose one module at a time to
complete all the book process (pdf’s in pieces). Then repeat the printing process until
complete all modules.
That’s all, your pdf’s will be freshly baked (with hard ingredients).

2.1.1. The Cook

We could “smell” in figure 5, the “fresh” watermark. This was a web process that is
seems to be involved in the moment when the content is printing, the cook put in there
the special element and we want to know which the recipe is. The cook is clever if don’t
want reveal nothing about it (putting it at the end of the process), but don’t know we could
be “cook of bits”.
Using the developer tools of our favorite browser, our magic word to invoke was
watermark.

We found a node called print_watermark_wrapper in the web code:

<body class=”fontsize_3 font_times textmargin_0 colormode_white”>
<div class=”dtbs_holder”>
<div class=”dtbs_book” style=”position: relative;”>

7
 <div class=”print_watermark_wrapper”></div>
We discovered where the “recipe” was hidden, we just needed get the ingredients and
start cooking.

2.2. Web App: Present

Not differs to much from the old version but some differences exist.
The print button is located at same position (left-bottom), the other print button is the
responsible to fire up the printing assistant directly, no pop-up new windows like in old
web version (section 2.1).

Figure 6: Printing functionality in new version

This second button isn’t necessary (has no sense in terms of usability), because there are
no other highlights/notes to print.
In next figure 7, we are showing the printing assistant from Google Chrome to see how
watermark differs.

8
 Figure 7: Printing functionality in new version

It’s less scattered in document, more vertical than horizontal in comparison as seen in
figure 5.

2.2.1. The Kitchen Assistant

Will be nice a little of help to know now where the “special ingredients” are.
We used again the developer tools to dissection the “evolutionistic recipe” and again,
same word to get “clues”.

Two matches were discovered, the node it’s also called: watermark

9
 Figure 8: Watermark node in new version

From the main node, we can see their children, all equal, with the same text repeated
several times distinguished as “watermark-tile”. So, it was discovered from its hiding
place again.

2.3. Native Application

Skillpipe borns with desktop and mobile “skills” but they announced the retirement of
the offline platforms by end of June 2019, their strategy is, focusing all efforts on one
application [6].
The offline apps are no longer available from the official channels, but we still keep the
desktop Windows Vista / 7 / 8 application. It will still work? Let's see through “the dust”.

2.3.1. Dusting It
We were excited to try it, we wanted to do a quick check and know if some process stops
at some point of the circuit. The desktop app is still installed, let’s click the .exe.

10
 Figure 9: Skillpipe desktop App - Initial screen

First thing eyes pointed was to the version, will be the latest? Appears to be the more up
to date under the Windows 7 edition [7]. It's odd that an offline app asks for a login, right?
If so, always a direct connection is mandatory, and maybe then you can reside offline
until the next re-execution or use the “Keep me signed in” (as seen in figure 9) to avoid
the credentials in every time.
The mouse was moving doing circles while thinking about it, and we find the “help”.
The app has a FAQ section that can be queried locally. Seems you need one-time
connection to internet to create and synchronize the profile, once done, reading offline
can be done. In next figure 10, it describes this process well in two sections under chapter
Skillpipe reader app for Windows Vista / 7 / 8.

11
 Figure 10: Skillpipe desktop App – FAQ

I had no plans to document this part, but the interest aroused in me to know how the
mechanism works to read the book using the credentials.

2.3.2. Caching It
I remember accessing to the app and download a book (in .bdb2 format) to start
interacting with. You can easily open the file with the corresponding extension (software
has the extension associated). The steps can be found in a YouTube video [8]: Access to
the Skillpipe Web App, then choose the book course, then download the file version and
open it. Finally, you need introduce again the credentials (on every app execution).
As a test, we tried open a .bdb2 we had, and login with our email as user but different
password to ensure if the connection against old Skillpipe desktop infrastructure is still
alive, and… Check your internet connection?

12
 Figure 11: Skillpipe desktop App – Login Unchecked

Internet connection at the moment of the test was fine! Let’s see what happen if try to
open the links Forgotten Password? and or click here to register now. Links are:
https://www.skillpipe.com/reader/en-GB/Account/ForgottenPassword ,
https://www.skillpipe.com/reader/en-GB/Account/Register respectively. Links appear to
be down, no longer exist (and at all moment, the Keep me signed in box is disabled).
Tried using the correct credentials and… Access granted! All the book functionalities
appear working OK, you can select the different modules and no error was found. The
profile should be store in some place of the computer…
In the config options, there is a function to retain the cache. It seems also contemplates
the credentials.

Figure 12: Skillpipe desktop App – Cache Options

The tree of folders and content it's as follows:

%USERPROFILE%\APPDATA\LOCAL\ARVATO\SKILLPIPE READER
\---cache

13
| aps.v2
| cd.v4
| ebm.v2
| n.v1
|
+---Books
| +---619fb4b7-e582-487d-9bc1-8fb028766fc4
| | \---7
| | | SI.v1
| | |
| | +---additions
| | | \---images
| | | *.png
| | |
| | |
| | +---html
| | | chap*.html
| | |
| | \---meta
| | additionalcontents.dat
| | meta.xml
| | thumbnail.png
| | toc.xml
| |
| \---7f09d2fa-bb31-48eb-9e48-2fb04b3aa562
| \---2
| | SI.v1
| |
| +---additions
| | +---images
| | | *.png
| | | cover.png
| | |
| | \---json
| | *.json
| |
| +---html
| | chap*.html
| |
| \---meta
| additionalcontents.dat
| meta.xml
| thumbnail.png
| toc.xml
|
\---Users
\---user___account
| ad.v1
| c.v1
| l.v2

14
 | p.v1
| ubi.v1
| us.v1
|
\---uc
+---619fb4b7-e582-487d-9bc1-8fb028766fc4
| ac.v1
| s.v1
| ucs.v1
|
\---7f09d2fa-bb31-48eb-9e48-2fb04b3aa562
s.v1
ucs.v1

Inside of SKILLPIPE READER folder, there is a cache folder with files which have a
non-related extension with a software (unknown). There are other important folders;
Books and Users but the extensions .v1, .v2, appear in, not only under cache folder, also
spread in the rest. The exception is with .v4 only is found in cache. All the .v* appears to
be encrypted (major of content is illegible).
The Books seems contains all the digital documents associated with the student. Folders
619fb4b7-e582-487d-9bc1-8fb028766fc4 and 7f09d2fa-bb31-48eb-9e48-2fb04b3aa562
are associated with 20331b-core-solutions-of-microsoft-sharepoint-server-2013-v7.bdb2
and 20462d-administering-microsoft-sql-server-v2.bdb2 books respectively. The images
located in images folder are not encrypted, neither files meta.xml and toc.xml but
chap*.html yes. Other folders from main Books folder, seems to be the document
structure.
Users folder contains other folder with the username as a name, with .v*, but the most
interesting is there are the folders of the books (same name) that are under Books folder,
in that case, only .v1 inside. One of them has one more .v1, the ac.v1 maybe it reference
to last read position.

Well, can’t spend time on reverse engineering, but one thing to try is copy the cache
folder and paste into other user account created in same computer to see if still works.
And worked! Next challenge was copy the contents but into other computer, and… it
fails! The error is the same as seen in figure 11 (and yes, the user and password was
correct!).
My idea was virtualizing the machine just in case some registry value or other file in
wherever were missing. Using this method, we could be sure that it is not about leaving
any files along the way.
The objective is converting it to a VirtualBox machine as destination. We used the
Disk2vhd from Sysinternals [9]. We unchecked the options of ‘use Vhdx’, and ‘Use
Volume Shadow Copy’ [10]. Conversion was OK, taking some long time but without any
error. Turned on and after a while, no difference from original Windows, so the clone was
almost perfect. Now, you want to know how was the result? Another fail!!

15
OKI Doki! So, the encryption mechanism is using something specific for every machine
(For what it seemed). And what about the inspection of a book in their BDB2 format?

2.3.3. BDB2
It consists, in see if there is any tool to dissection the book and 7-Zip was chosen for this
purpose.
For the test, we used the 20462d-administering-microsoft-sql-server-v2.bdb2. The
content is the same as seen in the tree of the chapter 2.3.2. the only difference is that the
file SI.v1 is not in the BDB2 skeleton.
I wanted to do a quick comparison taking the structure of an EPUB Before inspecting in
depth. Let’s show if BDB2 has some similitudes [11]:
| Mimetype file (Archive)
+---META-INF
| container.xml
\---DOCUMENT folder
contains HTML, CSS, image files, plus OPF and NCX files

Despite having a similar structure, Skillpipe seems to adopt other different mechanism
(plus something with DRM) although it seems to have good intention/inspiration with its
appearance in xml as the toc file (toc.xml from Skillpipe against toc.ncx from standard
EPUB).
In .bdb2, almost all the important content is encrypted like the html files. You can extract
other content, part of the book, but is like having “souvenir postcards”, if you don’t have
all the pieces, you never see the entire “picture” and many hypothesis should appear.
To be a collector of Skillpipe BDB2 books, your files are *.json, *.png and *.xml:
- JSON (\additions\json): files .json shows the Check Your Knowledge section.
There are many type of challenges (from the book), you need choose correct
answers or drag & drop…
- XML (\meta): meta.xml shows the description of the course. toc.xml the table of
contents (The chapters are related to the html files).
- PNG (\additions\images): All the .png images of the modules.

Nothing to do with encrypted books but… Why not look to the book without
dismembering it? Let's go for the magnifying glass!
We opened with notepad++ and we review first words at the beginning of the first line.
Three sections can be defined here.

16
 Figure 13: Opening/inspection a .bdb2 with notepad++

According to the figure 13, the first understandable letters are bdb2. This is the extension
that software will use to identify as their book.
The “strange” following letters could be something unique related with the license of the
student.
The number, 2 in this case, is the code of the update of the book, this code can be seen in
the tree, above in chapter 2.3.2 following the folder with name of the book id 7f09d2fa-
bb31-48eb-9e48-2fb04b3aa562. This can be also seen in the meta.xml file, as the
following figure 14 shows:

17
 Figure 14: Opening/inspection meta.xml with notepad++

It demonstrates number and id are used as identifiers.

The two last words, PK are not used when we have the books in cache. We created a
blank file with any name and extension .bdb2 with this content:
bdb2úÒ 1»ëHžH/°K:¥b 2
Open – put the credentials in Skillpipe, and it worked like a charm :D ! (Seems data is
consumed from the cache instead from the book file).
For curiosity, we could know what files are implicated from the login until the book is
open.

2.3.4. Tracking It
In order to know how the encryption mechanism is working, need some tool that helps
on this research, something that show us the accessed files by Skillpipe ordered by time.
If you work in this beautiful IT world, and you challenge with servers, networks… (like
a system administrator) You may know the great tools from NirSoft [12]. In all the variety
of tools NirSoft has developed, the suitable for this job is FileActivityWatch [13]. This

18
tool will help to display what files are accessed by Skillpipe in time order, and of those,
if is read, write, bytes on each action… Simple, good GUI and easy to track.

We started with the origin computer where is working well and then with the Virtualized
one.
Origin computer:
Before proceeding, need to prepare the filter correctly to avoid “noise” (because
FileActivityWatch gather all info of the files access in computer from all the processes).
In advanced options, the option can be set up. Now need filter the filter, we mean, capture
only in the moment of do a correct login avoiding the entire opening of the .exe-cutable.
To accomplish this, capturing can be stopped with F2 key, and if for some reason there
are “garbage” can be cleaned by Ctrl+X.
Time to click on Sign In Skillpipe button, and the following files has been accessed:

19
 Figure 15: FileActivityWatch – Filtering process in Advanced Options and File inspection

From top to down, the p.v1 is first file accessed preceded by the files c.v1 and us.v1.
Seems a check for the password, and these in conjunction a check with some computer
algorithm to be unique. Then at line 4 in same figure 15, log.txt file is touched and after
two more lines, is the turn of error.txt.
That is all? Why not take a look in logs of the application? The log folder is found in:
%localappdata%\arvato\skillpipelogs and from here, two text files (.txt). One is log.txt
and the other error.txt both maintain six more versions, in this prefix/format; log.txt.year-
month-day and error.txt.year-month-day.
A good idea could be correlate by time, so if FileActivityWatch say that at 11:56:52 PM
let’s see by this time, opening the log.txt (because the error.txt don’t show important
because all was fine!):

20
2020-01-29 23:56:52:5652 [5] INFO Remote logging disabled by user preference.
2020-01-29 23:56:52:5652 [1] WARN Calling SaveBookUserSettingsAsync when
UserContext.IsComplete = false
2020-01-29 23:56:52:5652 [7] INFO The operation 'Load book index' took '185' ms
2020-01-29 23:56:53:5653 [1] INFO The operation 'chapter loading' took '631' ms

The information we obtain is three INFO (information) and one WARN (warning). Good
news is book could load correctly, and take less than a second. The WARN appears to be
a function of the software that seems not to be more important.
Don’t get tired yet, there is another tool that could be great to use… is Skillpipe using
some port for connections (internally or externally)? Process Hacker [14] has “power
glasses” to say something about, and can filter by the application name:

Figure 16: Process Hacker – Network filtering

As note; the local and remote addresses are covered because is localhost, but shows the
FQDN of my test machine.
The port 39123 is open at the beginning (listen), and while book is open, there is a
bidirectional internal communication. Six connections are open. 39123 against 7316,
7317, 7318, 7319, 7320, 7321 and vice versa. It could open six connections to balance
the load or to distribute parts of functions in the program.
Maybe doing same operation in the cloned virtual machine more context could get, who
knows!

21
Cloned virtual Computer:
Let’s use again the FileActivityWatch with same steps described above. And the result is
shorten:

Figure 17: FileActivityWatch – Filtering process at virtual cloned machine

It pretends read and interpret the p.v1 file but seems without success because we see the
error.txt is first wrote than the log.txt. Next step? See what happen in error.txt (again,
correlate by same time):

2020-01-30 01:07:30:730 [5] ERROR Error occured while reading or deserializing a

file: C:\Users\user\AppData\Local\arvato\skillpipe reader\cache\Users\user\p.v1
System.Runtime.Serialization.SerializationException: Deserializing '? ??h*gf??"??{-
??A
o!z??????<??(?C?hZJ?m???-Ku3??Z????%Pu??T7?

22
?H?z?!l??•??;?D?(n?P/?m??B??r???=????????8?-
?? >?J•?•??#?????m? ?!6{?{?\??' failed --->
Newtonsoft.Json.JsonReaderException: Unexpected character encountered while
parsing value: ?. Path '', line 0, position 0.
at Newtonsoft.Json.JsonTextReader.ParseValue()
at Newtonsoft.Json.JsonTextReader.ReadInternal()
at Newtonsoft.Json.JsonTextReader.Read()
at
Newtonsoft.Json.Serialization.JsonSerializerInternalReader.ReadForType(JsonReader
reader, JsonContract contract, Boolean hasConverter)
at
Newtonsoft.Json.Serialization.JsonSerializerInternalReader.Deserialize(JsonReader
reader, Type objectType, Boolean checkAdditionalContent)
at Newtonsoft.Json.JsonSerializer.DeserializeInternal(JsonReader reader, Type
objectType)
at Newtonsoft.Json.JsonConvert.DeserializeObject(String value, Type type,
JsonSerializerSettings settings)
at Newtonsoft.Json.JsonConvert.DeserializeObject[T](String value,
JsonSerializerSettings settings)
at Arvato.DTBS.Common.Infrastructure.JsonNetJsonConverter.GKC`1.C()
at System.Threading.Tasks.Task`1.InnerInvoke()
at System.Threading.Tasks.Task.Execute()
--- End of stack trace from previous location where exception was thrown ---
at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
at
System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotific
ation(Task task)
at Arvato.DTBS.Common.Infrastructure.JsonNetJsonConverter.HKC`1.I()
--- End of inner exception stack trace ---
at Arvato.DTBS.Common.Infrastructure.JsonNetJsonConverter.HKC`1.I()
--- End of stack trace from previous location where exception was thrown ---
at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
at
System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotific
ation(Task task)
at System.Runtime.CompilerServices.TaskAwaiter.ValidateEnd(Task task)
at A.WP`1.TP.M()

Amount of words, more difficult to read (lot of errors). Basically, it fails to validate the
correct routines for p.v1 file (important lines are underlined in code).
Is important look for the exact times, but if we see the before log time? Let’s see so:

2020-01-30 01:07:09:79 [5] ERROR Error occured while reading or deserializing a file:
C:\Users\fuguet-gerard\AppData\Local\arvato\skillpipe reader\cache\n.v1
System.OverflowException: Arithmetic operation resulted in an overflow.
at Arvato.DTBS.Crypto.Implementation.BlowfishCrypter.Decrypt(Byte[] key, Byte[]
dataToDecrypt)
at Arvato.DTBS.Crypto.Implementation.BlowfishCrypter.G.E()
23
 at System.Threading.Tasks.Task`1.InnerInvoke()
at System.Threading.Tasks.Task.Execute()
--- End of stack trace from previous location where exception was thrown ---
at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
at
System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotific
ation(Task task)
at Arvato.DTBS.Common.Core.Crypto.MachineSpecificCrypter.MDC.H()
--- End of stack trace from previous location where exception was thrown ---
at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess(Task task)
at
System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotific
ation(Task task)
at System.Runtime.CompilerServices.TaskAwaiter.ValidateEnd(Task task)
at A.WP`1.TP.M()

It is talking about Blowfish algorithm designed by Bruce Schneier [15]. And about
MachineSpecific… So, the algorithm uses something of your machine to be able to
decrypt the content. File n.v1 appears to be “an essential key” in conjunction with the
machine/computer to achieve the “clear work” of decryption.
We better “not bothering the blowfish”, lest we get poisoned!

3. Water-Engine-Mark
The native application does not work; it is not our mobility solution, not only because we
cannot move freely between devices, but because printing seems impossible...
Returning to the watermark; it’s digital and hides the content using an steganographic
technique. Aims the intellectual protection of the digital object (copyright). [16]
We are going to simplify the process adapting it to the specific case of the application,
we will not deepen much in all its aspects, but some components need to be exposed,
must be explained to understand its mechanics.

3.1. Pipe Factory

Let’s show first a diagram and going to explain the legend:

24
 Figure 18: Watermarked Process

The pipe simulates the marked of the process (where watermark is made), is like the
processes of a factory:
- (O)bject: Is the document to be subjected to stamping.
- (K)ey: Is who let you access to the marked of the process and to the
object/document. In this case the key are the student credentials to access to
Skillpipe portal. Without the credentials, no object (nor stamped).
- (S)tamp: Is the watermark to be inserted in the object/document.
- (SO)stamped object: Is the result of (O)bject & (S)tamp, it means, it produces the
object with the watermark.

3.2. Types
There are different watermark styles (it depends on the object and if it is hidden or
uncovered), for those, we will identify what Skillpipe uses:
- Private: Original (O)bject is needed
Type 1: SO x O x K → S
Type 2: SO x O x K x S → Stamped/NoStamp
- Semi-private: The (S)tamp is needed
SO x K x S → Stamped/NoStamp
- Public: Not need the original object (O) neither the stamp (S)
SO x K → S
- Visible: The watermark here is visible, through the key (K), the original object
can be retrieved without losses. The stamp (S) can be pulled off as if it were
another layer
SO x K → O

After view all the types, is easy to identify the case of the Skillpipe, visible type is the
correct formula, because with the credentials, you have access to the material and the

25
content can be seen without the watermark. The problem still exists (and persists) at the
printing stage.

4. Pipe Leak
Have you ever had a leak in a pipe? Well, it depends on the home and the age of it or
quality of pipe materials, but surely, you have already seen it in the street (water pipe as
most typically). No one would think of breaking one of them, because the disaster that is
created is very scary! But…! the pipe of Skillpipe “is dirty”, we need to break it to repair
and clean it.

Figure 19: Stripping the watermark

If we interrupt the process of watermarking, we will able to catch the original object (O)
without being stamped (S). This is the theory, the work plan, now, we must see how to
put it into practice.

4.1. Manual Repair

This is the procedure of the hack (make the watermark disappear) deleting manually the
node that correspond to watermark code in web. We will show the old and nowadays
mode to achieve this.

4.1.1. Cast Iron (Old)

I used this method in 2017 so we can say is little old and is no longer available, but for
historical reasons, we put here.

26
Step 1: Have you the (K)ey? Enter to the portal.
1. Go to skillpipe.com and enter the credentials.

Figure 20: Skillpipe – Sign In

2. Choose the course you want to keep offline.

Figure 21: Skillpipe – Choosing Course

Step 2: Take out the water from the Mark.

1. Choose a desired module to print.

27
 Figure 22: Skillpipe – Choosing Module

2. At the bottom – left, you can localize the print button, click and again click to print in
the window that appear to the right (seen in figure 4 of the section 2.1.).
3. A new tab is created, and the print wizard appears, cancel the print wizard window
(seen in figure 5 of the section 2.1.).
4. Access to the developer tools of the browser (we used Firefox) options pushing F12
and look for the node: print_watermark_wrapper. Inspector option can be helpful.

Figure 23: Deleting the Watermark through Developer Tools

5. Right click – Delete Node.

Step 3: Get the unstamped object.
1. You can invoke again the print wizard with Ctrl + P and you will be surprised that the
watermark is gone.

28
If you print following the process until Step 2 at point 3 without cancelling the print
wizard and without deleting the print_watermark_wrapper node, you can compare the
two documents, and you be realized the size of the non-watermarked file is too heavy!!
A video of all the process was made, here you have the link (in Spanish but with English
subtitles):
SkillPipe - Eliminar Marca de Agua (Lector de Cursos usado por Microsoft)
https://youtu.be/wqXxGh3D-CI

4.1.2. PVC (New)

A very similar method to the previous one but with two differences; the watermark node
is somewhat deeper in the code and, there is no new tab to show you the result of the
“finish of the product”. Print wizard appears in the same tab and you must have “faith”
at the end of the printing process… Or you can use the Chrome browser to see if all is
going well (viewing print preview)!
0. Let’s see the version of the new Web App first.

Figure 24: Skilpipe Web App Reader - About

Repeat all the steps of the step 1, step 2 until complete the pass 1. We are going to continue
in pass 2 of the step 2 (seen on section 4.1.1.).
2. Before printing, the watermark node must be cleaned/deleted. Open DevTools of the
browser (we used Chrome now) options pushing F12 and search for the node:
#watermark. Then Right click – Delete element as shown in the following figure 25.

29
 Figure 25: Chrome – DevTools – Deleting watermark node

Step 3: Drink water without mark.

1. You can close the DevTools window and print using the Web App print button (seen
in figure 6 of the section 2.2.) and you will aware, watermark is not shown in document!

Figure 26: Chrome – Print Wizard – Watermark is gone

30
4.2. Semi-Auto Repair
The manual method of the “pipe repairing”, works well, watermark is removed entirely
in all the book module, however, you need repeat the process on each module and can be
tedious (It is already quite tedious to generate a PDF for each module…). So decided
thinking in something, some kind of script. I still remember when the teacher says that a
script at the beginning consumes you a significant amount of time but, when is build, your
working life turns a little easier. Need you always create one? Not always. A workmate
who has very good skills in development, before creating a new software, he searches if
someone created it. Why do the job twice? This is we did, and we found a software able
to defeat the watermark.

4.2.1. Ad Blocking
An Ad Blocker is used usually as a plugin of a browser to defeat the advertisings. Not
only can block advertisings, can also filter and not show other things, things like
watermarks. The PoC was done with ABP (Adblock Plus) [17]. Let’s see how it works!
0. Showing the version of Chrome browser & ABP used in this PoC.

Figure 27: Chrome & ABP versions respectively

Step 1: Imprisoning to the watermark.

1. Open the ABP icon located at top – right of Chrome and click on settings button.

31
 Figure 28: Opening ABP plugin in Chrome (Incognito mode)

2. Go to Advanced – Type the following filter in My Filter List:

skillpipe.com###watermark [18] and push +ADD to add it.

Figure 29: Opening ABP plugin in Chrome (Incognito mode)

Step 2: Ready to print clearly.

1. All is ready to print without the watermark, so situate in the module of a course, and
print it with the printing functionality of the Skillpipe Web App (seen in figure 6 of the
section 2.2.).
2. That’s all! You can print next modules without the watermark.

32
This new method is covered in a video showing all the process (including the new manual
method seen in section 4.1.2.). Here you have the link:
Skillpipe - Remove the Watermark (Course Reader used by Microsoft) using Ad Blocking
https://youtu.be/Ok1UmRFWoLY

5. Conclusions
Watermark in digital content is still fragile and lacks robustness.
In 2017 we initiate the notify, we reach at the webform of the Skillpipe official support:

Figure 30: Contacting with Microsoft Training & Certification Support

And we got an answer from the Arvato Bertelsmann staff (by Arvato Distribution GmbH,
an independent reseller of Microsoft) [19]:

33
 Figure 31: Answer of Microsoft Courseware Fulfillment Center

In addition, they are saying this is normal behavior so, no changes?! Well… It is clear
that the answer did not convince me at all. Like in Windows OS’s, for do some operations
you can achieve it by many ways, this is we did, and another email went to the MSRC
[20]:

34
 Figure 32: Email request to Microsoft Security Response Center

We attached screenshot and attached the email response evidence from Courseware,
including the ticket number, … The answer came:

Figure 33: Email answer from Microsoft Security Response Center

Issue is not related with Microsoft product, so again go to Skillpipe was our scape, but we
entered in an infinite loop… until we decided to stop it, doing this white paper, with good
intention (always) of course.

35
Is true Skillpipe developers made some changes in code (comparison is in section 2.1.1.
and 2.2.1.) but again they were insufficient to protect against total disassembly of the
watermark.
As I mentioned in section 1. The initial problems and objective were trying to read the
books/courses in a comfortable way, avoiding watermark problems (and we aren’t against
the copyright). The mechanism to protect these courses is immature, and probably we
wouldn’t try to break or clean it if the watermark had the characteristics like; be
imperceptible and mustn't degrade the object intended to protect. The comparisons with
and without the watermark are enough evidences to address a solution to the
watermarking problem, but Google Chrome seems to know how to handle the problem
of the weight of PDF, vertical bars in the corners and text selection. This implied us to
test using the PDF Creator [21] (used with Firefox combination at the beginning, in 2017)
with Chrome and the built-in PDF that Firefox doesn’t have [22] (only through add-ons).
The following table show the results in bytes for each case (we did the tests with and
without watermark printing the first Module 1 of the SQL course):

Size on PDF PDF Creator 2.5.2 Built-in PDF

in Bytes
Firefox Chrome Chrome
Watermark 5.969.265 3.658.241 830.608
No Watermark 594.228 688.974 821.876

Figure 34: Comparison table of PDF sizes

Note that under watermark the clear winner is Chrome but in “No watermark” the best
results are taken by Firefox (which uses PDF Creator) sizing 581kB against 803kB in
Built-in PDF. Even so, PDF Creator appears to be better algorithm working without the
watermark, Chrome resized to 673kB.
About the vertical bars and the selection text, not all the culprit is for PDF Creator.
Firefox manipulates it in one way and Chrome in another different. See the differences at
the following figure 35:

36
 Figure 35: Comparison with Firefox (at left) and Chrome (at right) using PDF Creator

In Firefox version, the vertical bars and text distortion persists along the module and not
all the text can be selected. On the other hand, in Chrome, all is “smoothly” including the
size of the PDF file, indicating the watermark is more text than a raw image.

As final words to add for the end… How can we strengthen the watermark?
Most important is take care of the integrity of the digital document, must be robust and
secure but not produce a high distortion of the object. We saw some tools are useful for
developers (like the DevTools of the browsers) but also are dangerous allowing tamper it.
Taking this into account, some method in code could do the job.
We took out the watermark before producing into PDF. So, PDF here is not the key
element to put into check, maybe a review on the mechanism used could be good, or the
creation of another format, be subject to another object, etc.

And other question could be, is the watermark the right way to protect the documents
against illegal copies, should it be abandoned in the digital world and keep using in the
physical?

Either way, this document is free of any watermark. Keep it clean!

Remember…

37
Be Good, Be Hackers.

6. References
[1] Computer Training, Computer Certifications, Microsoft Learning.
https://www.microsoft.com/learning
[2] Skillpipe eReader. https://skillpipe.com
[3] Skillpipe - Support for the offline apps has ended.
https://policies.skillpipe.com/en/faq/#news
[4] Skillpipe – About this app. https://www.skillpipe.com/#/account/about
[5] SkillPipe Frequently Asked Questions – Directions Training Center.
https://directionstraining.zendesk.com/hc/en-us/articles/200606785-SkillPipe-
Frequently-Asked-Questions
[6] Skillpipe Mobile - From Native App to Progressive Web App.
https://pmskillpipeweu.blob.core.windows.net/archive/Skillpipe%20Mobile_From%20
Native%20App%20to%20Progressive%20Web%20App.pdf
[7] Skillpipe Reader - Software Informer. https://skillpipe-reader.software.informer.com
[8] YouTube - Training & Education Services - Download A Book.
https://youtu.be/foSx0pi4uS8
[9] Disk2vhd - Windows Sysinternals. https://docs.microsoft.com/en-
us/sysinternals/downloads/disk2vhd
[10] How to Convert Physical Windows Computer into a VirtualBox Image - Jozef
Jarosciak Blog. https://www.joe0.com/2017/09/27/how-to-convert-physical-windows-
computer-to-virtualbox-virtual-machine/
[11] Learning About EPUB: Structure and Content - Altova Blog.
https://www.altova.com/blog/learning-about-epub-structure-and-content/
[12] NirSoft - freeware utilities: password recovery, system utilities, desktop utilities.
https://www.nirsoft.net/
[13] FileActivityWatch - View read / write / delete file activity on Windows 10/8/7/Vista.
https://www.nirsoft.net/utils/file_activity_watch.html
[14] Overview – Process Hacker. https://processhacker.sourceforge.io/
[15] Schneier on Security: The Blowfish Encryption Algorithm.
https://www.schneier.com/academic/blowfish/
[16] Wikipedia - Digital Watermark (Spanish).
https://es.wikipedia.org/wiki/Marca_de_agua_digital
[17] Adblock Plus - The world's # 1 free ad blocker. https://adblockplus.org

38
[18] How to write filters - Adblock Plus Help Center .
https://help.eyeo.com/adblockplus/how-to-write-filters
[19] Courseware Marketplace. https://shop.courseware-marketplace.com
[20] MSRC - Microsoft Security Response Center. https://www.microsoft.com/msrc
[21] Create, edit and merge PDFs easily - pdfforge. https://www.pdfforge.org
[22] How to print a page on firefox in pdf - Firefox Support Forum - Mozilla Support.
https://support.mozilla.org/en-US/questions/1109428

39