Pandora’s Email Box? An Exploratory Study of Web-Based Email Forgery Detection and Validation.

Journal of Digital Forensics, Journal of Digital Forensics,

Security and Law Security and Law

Volume 7 Number 1 Article 3

2012

Pandora’s Email Box? An Exploratory Study of Web-Based Email Pandora’s Email Box? An Exploratory Study of Web-Based Email

Forgery Detection and Validation. Forgery Detection and Validation.

Richard Boddington

Murdoch University

Grant Boxall

Murdoch University

Jeremy Ardley

Dalton Pty Ltd

Follow this and additional works at: https://commons.erau.edu/jdfsl

Part of the Computer Engineering Commons, Computer Law Commons, Electrical and Computer

Engineering Commons, Forensic Science and Technology Commons, and the Information Security

Commons

Recommended Citation Recommended Citation

Boddington, Richard; Boxall, Grant; and Ardley, Jeremy (2012) "Pandora’s Email Box? An Exploratory Study

of Web-Based Email Forgery Detection and Validation.,"

Journal of Digital Forensics, Security and Law

Vol. 7 : No. 1 , Article 3.

DOI: https://doi.org/10.15394/jdfsl.2012.1111

Available at: https://commons.erau.edu/jdfsl/vol7/iss1/3

This Article is brought to you for free and open access by

the Journals at Scholarly Commons. It has been

accepted for inclusion in Journal of Digital Forensics,

Security and Law by an authorized administrator of

Scholarly Commons. For more information, please

contact [email protected].

(c)ADFSL

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Pandora’s Email Box? An Exploratory Study of

Web-Based Email Forgery Detection and

Validation.

Richard Boddington

School of IT, Murdoch University

Perth, WA 6150, Australia.

[email protected]

Tel: +61 893602801 Fax: +61 89360 2941

Grant Boxall

School of IT, Murdoch University

Perth, WA 6150, Australia.

[email protected]

Tel: +61 893602801 Fax: +61 89360 2941

Jeremy Ardley

Dalton Pty Ltd

Forensic Information Analysts

[email protected]

Tel: +61 8 6280 0008

ABSTRACT

Web based email systems may be a source of pristine digital evidence because of

the perceived difficulty of client tampering with messages stored inside the email

account. We demonstrate that such assumption is wrong in the case of Windows

Live Hotmail

®1

. Windows Live Mail

synchronises message on client-side

computers with the Hotmail

server, benefiting users wishing to synchronise their

email accounts and personal devices. However, this synchronisation opens an

exploit for wrongdoers to tamper with existing email messages and attachments as

well as facilitating the insertion of fabricated messages. The exploit process

enables persistent storage of tampered and fabricated messages on the Hotmail

®1

server. The exploitation favours both account owners and wrongdoers who gain

unauthorised access of others’ accounts. Even if tampering were suspected, we

anticipate some difficulties in validating messages to determine their reliability

and relevance. We predict, with trepidation, that the exploit process will become

commonplace and pose greater challenges to the cyber forensics examiner and

Hotmail

, Windows Live Hotmail

and Windows Live Mail

are the registered

trademarks of Windows Corporation.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

legal practitioner during investigations and legal proceedings. Regrettably, the

exploit complements the existing arsenal of tools for email forgery. More

ominously, it provides opportunity for traceless injection of illicit

material/malware onto any machine synchronised with the Hotmail

account.

Keywords Digital evidence, evidence validation, Windows Live Mail

, email

tampering, web-based email exploitation.

1. INTRODUCTION

Covert attacks to gain control over other users’ web-based email accounts for a

range of illegal and unethical purposes is not a new or uncommon phenomenon

(Florencio & Herley, 2007). Use of email systems to promulgate the spread of

malicious software capable of breaching privacy, disabling individuals’

computers and networks, and a myriad of scams, are unwelcome but well-

entrenched phenomena (Sunner, 2005). Bogus email messages created with little

technical skill can override email identity checking process, providing anonymity

for the miscreants and when delivered can have disastrous outcomes for victims

of such ploys (Levi & Koc, 2011). A significant vulnerability is poor password

security measures used by email providers, aggravated by weak user passwords,

which in turn facilitates, if not actually encourages exploitation of this essential

communications medium (Craddock, 2011; Preibusch & Bonneau, 2010).

The ability to access others’ email accounts allows intruders to create, delete,

transmit, move and copy messages but little else. An intruder, or account holder

wishing to modify an existing email message for some improper purpose may be

able to export messages, modify them but then find it impossible to reinsert the

emails into web-based accounts. It was considered difficult, if not impossible, to

modify web-based email messages stored on vendors’ servers without direct

access to the server by means other than the web page (Ardley, 2011).

We became aware of a current criminal case

during which the defendant

suggested the possibility of the complainant tampering with messages received by

the complainant from the defendant in a Hotmail

web-based email account. The

suggestion was this was done as a means to implicate the defendant in a criminal

activity. Initially, the proposition seemed improbable because of perceived

technical difficulties in editing message content and was dismissed by the

prosecution team of cyber forensics experts as being technically beyond the

ability of the average home computer user without advanced programming skills.

Nevertheless, the defence, forensics team considered it was possible with an

unknown but probably low level of difficulty, and further research would help to

identify and test simple processes allowing authorised and unauthorised

tampering of Hotmail

messages to succeed.

While the case is sub judice we are not permitted to identify the court or parties

involved.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

In this paper, we show that Windows Live Hotmail

messages can be modified

with a modicum of skill and that simple processes do exist to overcome

export/import issues as well as obliterating traces of modification used in these

processes. We observed that it was a relatively simple process to access and

modify Hotmail

messages by a non-technical forger using Windows Live Mail

(WLM); an unannounced vulnerability we contend existed since 2007. The

processes we tested confirmed that it in some instances, it was possible to produce

near perfect forgeries. We will demonstrate that a Windows Live Hotmail

account can be synchronised with WLM by the account holder but unlike the

Hotmail

account, WLM can be used to alter existing messages and insert

fabricated messages into the Hotmail

account stored on the Hotmail

server.

Reliance on Hotmail

messages as unadulterated digital evidence is questionable,

and confirmation of the WLM exploit means that some form of validation is

required. If tampering evidence on the forger’s computer were purged, it is likely

that the Hotmail

server would provide the only possible means to detect and

verify message tampering and fabrication. However, reliance on Internet Service

Providers to provide full historical records of their client’s Hotmail

communications may prove disappointing to law enforcement agencies seeking

confirmation of tampering and fabrication because of insufficient message

logging. We note, for example that in contrast to its European and American

counterparts, Australia does not require Internet Service Providers to maintain

detailed logs of its users’ Internet activities including email messaging. Australia

does not presently possess powerful legislative standards such as the European

Telecommunications Standards Institute requiring service providers to retain

significant sets of data on their clients that assist law enforcement in investigating

crime and seeking exculpatory evidence to eliminate the innocent from their

investigations (Attorney-General’s Department, 2010).

Australia’s Telecommunications (Interception and Access) Act 1979 permits

service providers to provide telecommunications data, including email data, to

law enforcement agencies and telecommunications data and related information

kept for billing and other business purposes (Attorney-General’s Department,

2010). Australian federal and state government laws do not mandate service

providers to retain Internet data (e.g. SMTP records or mail client access other

than authentication) and email providers are unlikely to record email logs for any

length of time or hold sufficient data to validate suspect email messages. This

apparent lack of data available from service provider logs to assist examination

was further incentive to undertake our study to see what other evidence of

tampering the WLM exploit might provide.

Notwithstanding legislation empowering law enforcement agencies to obtain

email records from email vendors in criminal investigations, it is not always so

straightforward in civil cases. In a 2008 civil trial (Alexander, 2008) a

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Mississippi district court upheld the defendants’ right to seek the quashing of a

court order by the other party seeking detail of the email accounts of the

defendants’ employees. The court ruled that the relevant statute the third party,

the email vendor, may not disclose such information in civil matters (Alexander,

2008).

We foresee that evidence obtained by forensics examiners from Windows Live

Hotmail

server logs that hitherto provided authoritative views on message

antecedents and attributes might no longer be relied upon per se. Although email

communications exploitation through a broad range of attacks has been present

since its creation, we anticipate that WLM, while offering many benefits to its

users, may be one of a number of applications inadvertently nurturing email

forgery in a variety of forms.

Although at the time of writing were unable to find any scientific literature on the

WLM vulnerability or other exploits, we thought it prudent to publish and make

aware those responsible for law enforcement and the courts that the integrity of

web-based emails should not be taken at face value. We suggest that some means

of validating email messages be applied when circumstances dictate or ideally, as

a matter of standard forensics practice. We have also engaged with Microsoft via

appropriate channels and received a reply that stated they were aware of the

functionality of the application but that it was essential to, “. . . provide a more

complete service to customers wanting email synchronized across multiple

devices”, but Microsoft was unaware of the exploitation process per se (Ardley,

2011, pp. 3-4).

Of concern to us was whether email messages could be tampered with and

fabricated messages could be inserted into a target Hotmail

account using WLM.

If so, certain criteria must establish alteration of the message content, the message

headers and the message attachments as well as insertion of fabricated messages.

It was necessary to establish that these changes would persist in the Hotmail

account linked to the server, not solely in the WLM account after an extended

period

. To establish a proof of concept, we required certain conditions that would

confirm conclusively the WLM exploit process, namely:

1. Message content could be altered.

2. Message headers could be altered.

3. Message attachments could be altered.

4. Fabricated messages could be inserted.

This paper looks specifically at the current version of Windows Live Hotmail

and the vulnerability of email messages for tampering. Time did not permit us to

An arbitrary eight week-period would demonstrate stability and

persistence of the tampered messages on the Hotmail

sever over an

extended period.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

undertake similar, detailed examinations using other web-based email

applications, notably Gmail™, Yahoo 7 Mail

, AOL Mail

or Outlook

, but the

results observed from exploratory examinations of these applications leads us to

suspect that they too were vulnerable to tampering via WLM, and in the case of

Outlook

, through the simple ‘drag and drop’ migration process.

In this paper, we outline the history and nature of Hotmail

, Windows Live

Hotmail

and WLM. We define and describe the process of the WLM exploit and

the extent to which it can be used to modify Hotmail

message content, headers

and attachments. We describe how we tested the proof of concept and the

outcomes, and how we compared the modifications with original email data. We

also highlight the challenges to validating messages facing the cyber forensics

practitioner

2. SUMMARY OF HOTMAIL™ AND WINDOWS LIVE MAIL

Hotmail

, one of the pioneering web-based emails, was made available to the

public free of charge in 1996 (Craddock, 2010a). In 2004, Hotmail

was moved

onto a system using Windows Server and Windows SQL Server

and more

recently upgraded to the latest version of SQL server (Craddock, 2010a). In 2007,

Microsoft released a beta version of its free email application WLM to replace

Outlook Express

on Windows XP

and Windows Mail

on Windows Vista

WLM incorporated the DeltaSync

®5

protocol enabling users to synchronise

Hotmail

and other email accounts with WLM (LeBlanc, 2007a, 2007b; Sierra,

2010). By 2010, Hotmail

accounts synchronised with WLM offered users

synchronicity between their PC email client, their browser, and their phone

(Craddock, 2010b). There is evidence that Outlook

has had a functional

synchronisation for some time, perhaps preceding 2007 using WebDav (LeBlanc,

2007b).

When installed on a user’s computer under Vista

or Windows 7

, WLM

typically creates a default directory named Windows Live Mail under the

Microsoft folder in the nominated user’s Users folder. Users are able to add email

accounts by using the menu feature in WLM, provided they add the email account

name and password. Folders under Hotmail

accounts are created in WLM, most

usually, Inbox, Draft, Sent items, Junk email and Deleted items. This is very

similar to some IMAP email-clients such as Thunderbird that provide similar

directory structures and individual files.

Gmail is owned by Google Incorporated. Yahoo 7 Mail

is owned by Yahoo. AOL

Mail is owned by AOL Incorporated.

Outlook

is a registered trademark of Microsoft Corporation.

These applications and operating systems are registered trademarks of Microsoft

Corporation.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

3. THE PROCESSES INVOLVED IN THE WLM EXPLOIT

Unlike its predecessors, Outlook Express

and Windows Mail

, which stored

messages in less accessible formats and obscure folders, WLM folders are readily

visible using Windows Explorer, for example, and each message can be opened,

viewed, modified and saved using Notepad, and other text editing software. More

remarkable is that when WLM is running, it is also possible to drag and copy

email messages from any of the message folders to the desktop, edit the message

in various ways, and drag the message back into WLM. This feature is not

obvious to users and we do not believe it was intended to be part of the normal

use of the application. It seemed likely that tampering of email messages could

occur because of the way WLM stored messages from Hotmail

accounts on the

client machine and these messages would be migrated to the vendor server. If so,

these changes were expected to persist server-side and may assist in detecting

tampering.

The account holder may access the Hotmail

account either by direct access to the

Hotmail

account held on the server or by installing WLM and synchronising

with the Hotmail

account as shown in Figure 1a. Opening the Hotmail

account

by web access does not permit the user to tamper and insert a fabricated message

into the account. However, accessing the Hotmail

account through WLM

facilitates exploitation and allows the account holder to alter existing sent and

received messages for a variety of reasons.

Figure 1b presents an alternative scenario is when an attacker gains access to the

account holder’s computer and uses WLM to access the Hotmail

server and

tamper with or insert messages.

A remote attack is shown in Figure 2 where the attacker has the account holder’s

account name and password. The attacker is able to access the Hotmail

server by

synchronising through WLM and tamper with and insert messages.

Outlook Express

, Outlook 2011

and Windows Mail

are registered trademarks of

Microsoft Corporation.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Figure 1a. The account user accesses the Hotmail

server by direct access to

the email account or uses WLM to synchronise and access the account.

Figure 1b. An attacker gains access to the account holder’s computer to

tamper with or insert messages.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Figure 2. An attacker on a remote computer using WLM to synchronise with

the account holder’s account on the Hotmail

server to tamper and insert

messages.

Figure 3 illustrates how WLM on an account holder’s computer synchronises

with the Hotmail

account held on the Hotmail

server. The Hotmail

account is

added to WLM using the account name and password. The WLM account is

opened on the account holder’s computer and a message selected for tampering is

dragged from WLM by the ‘drag and drop’ facility and placed on the desktop and

the original message in the account is deleted to conceal the forgery. The

extracted message may be edited using a text editor such as Notepad then saved

and dragged back into the account in WLM. The account is manually or

automatically synchronised with the Hotmail

server and the tampered message

remains in the Hotmail

account.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Figure 3. The sequence of events showing an account holder altering an

existing email message.

Similarly, in the event an attacker hijacks the account holder’s computer or

initiates an attack from another computer, the attacker has the ability to access the

account holder’s account through the WLM exploit either on the account holder’s

computer or on the attacker’s computer, as shown in Figure 4. The process of

message tampering and insertion is identical with the process illustrated in Figure

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Figure 4. The sequence of events showing an attacker altering a message in

account holder’s account.

In the first two instances, evidence artifacts of tampering are confined to the

account holder’s computer and the Hotmail

server. In the remote attack,

evidence of tampering would be located on the attacker’s computer and the

Hotmail

server; server-side is dependent on appropriate server logs and retention

periods of those logs.

We discuss the issues of evidence location and characteristics in Section 8.

4. APPLICATIONS USED AND PROCESSES UNDERTAKEN TO TEST

THE PROOF OF CONCEPT

For the proof of concept, we used a desktop computer running Windows 7 Home

Premium

operating system and Windows Live Mail 2011

email application

For ease of reference, we called this computer the ‘editing computer’. A second

computer, the ‘independent computer’, compared and checked the outcomes

obtained by the ‘editing computer’. Identical operating system and email

Windows 7 Home Premium

operating system is the registered

trademark of Microsoft Corporation.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

application were installed on the ‘independent computer’ as were used on the

‘editing computer’. The use of the ‘independent computer’ would ensure that

results observed on the ‘editing computer’ could be verified independently. The

computer mouse was used to transfer twelve candidate message files (.eml) from

the Hotmail

account in WLM onto the ‘editing computer’ desktop and vice versa

using a ‘drag and drop’ technique. Notepad was used to edit the extracted files

and message files in the WLM folders present on the ‘editing computer’ directory.

A Windows Live Hotmail

account was created using the ‘editing computer’ for

testing the processes and efficacy of message tampering. The account was

populated with messages sent from other Windows Live Hotmail

, Gmail™,

Yahoo 7 Mail

, AOL Mail, and POP3/SMTP accounts created specifically for the

study on a separate computer. Email messages were sent from the Hotmail

account on the ‘editing computer’ to the other email accounts. These processes

would enable later study of the characteristics of the tampered messages and

fabricated messages located in the respective Inbox and Sent items folders on the

‘editing computer’.

WLM was installed and opened on the ‘editing computer’ where tampering and

fabrication of messages would occur. The Hotmail

account previously created

for the purpose of the study was synchronised successfully with WLM on the

‘editing computer’ and we noted confirmation during the account

synchronisation, that the mail server was an HTTP server implementing

DeltaSync

version 2.0.0. We observed the same version of DeltaSync

was

installed during the installation of WLM on the ‘independent computer’.

The study commenced with attempts to copy and modify email messages in the

Inbox and Sent items folders on the ‘editing computer’ by dragging twelve

prepared messages from each folder to the computer desktop to see whether

tampering of the message content, headers and attachments were possible.

Notepad was used to change headers and message content and then the messages

were reinserted into the respective message folder, in some instances with the

original message extant, in others with the message previously deleted.

Observations would determine whether these processes resulted in persistence of

the migrated file remaining in each message folder. On completion of each

tampering process, hashes were taken of each file for later comparison during our

study into the persistence of the messages in the Hotmail

account and the

Hotmail

server. WLM was closed and the computer rebooted to determine

whether the tampered messages persisted in the folders.

We then studied the viability of tampering with extracted messages by

substituting original image and text file attachments with previously prepared

messages to facilitate the substitutions on the ‘editing computer’. We used

Notepad to alter and insert the fabricated scripts into original messages, two in the

Inbox folder and two in the Sent items folder. Once the substitution was

completed, the messages were reinserted into their original folders onto the

Journal of Digital Forensics, Security and Law, Vol. 7(1)

‘editing computer’ using the ‘drag and drop’ process. We then manually

synchronised WLM with the Hotmail

server.

To test whether fabricated messages could be inserted into WLM and

synchronised with the Hotmail

server, we used twelve messages from a different

Hotmail

account which we dragged to the ‘editing computer’ desktop then

inserted them into the Inbox and Sent items folders of WLM, which was then

synchronised with the server. Four of the messages contained image file

attachments and the remaining four messages contained text file attachments.

We also wished to observe whether the tampered and fabricated messages

inserted into the WLM folders would synchronise and remain on the Hotmail

server. WLM was manually synchronised after each insertion and the ‘editing

computer’ shut down and rebooted, WLM was opened and observations made of

the presence and characteristics of the tampered and inserted messages to

determine their persistence in WLM. This process was repeated by using a WLM

account on the ‘independent computer’ to see whether the tampered message

would populate WLM and to observe the characteristics of the messages. To

determine whether the files had synchronised with the Hotmail

server, we used

the ‘editing computer’ and the ‘independent computer’ and opened the Hotmail

account to see whether the tampered files were present and observe their

characteristics. We assumed that an eight week-period would be sufficient to

confirm that the tampered files persisted on the Hotmail

server.

To gain more information about the DeltaSync

synchronisation process, namely,

the characteristics of the message metadata recorded on the Hotmail

server, we

simulated conditions of the DeltaSync

synchronisation as closely as possible.

DeltaSync

is not freely available for use separate from Microsoft products

and

we were unable to identify the type of metadata stored on the Hotmail

server.

Consequently, we used the UNIX utility rsync as the nearest known

approximation to DeltaSync

to simulate and investigate the server-side actions.

The X-Ways Forensics

analysis tool was used to locate and examine the folder

and email message files and obtain file hashes for comparison and event

reconstruction purposes. This provided confirmation of changes to tampered

messages and proof of reinsertion into WLM and the Hotmail

control account;

physical examination of the computer directories providing more complete file

attributes and antecedents.

Forensics images were taken of the ‘editing computer’ after the tampering was

completed and the WLM synchronisation with the Hotmail

server completed, to

observe and obtain the characteristics of the WLM folders and messages for signs

Note: DeltaSync

has been reverse engineered at the client side.

X-Ways Forensics is the copyright of X-Ways Software Technology

AG.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

of tampering. Searches of tampering events and migrated message files were also

made to see if any typical evidence artifacts were available to the forensics

examiner.

These tests were replicated using a ‘third computer’ with Windows XP

with

Service Pack 2 installed and a compatible version of WLM. This would allow us

to compare results obtained on the ‘third computer’ with the ‘editing computer’.

Section 5 provides details of each examination and our observations so that the

reader may also repeat the examinations themselves.

5. TESTS OF WLM BEHAVIOUR DURING MESSAGE

TAMPERING

Question 1: Do tampered and inserted messages synchronise with the

Hotmail

server and persist server-side for an extended period?

Test 1A: Eight tampered messages and four fabricated messages were

created on the ‘editing computer’ and were inserted into

WLM then synchronised with the Hotmail

server. The

messages were hashed for later comparison and WLM was

closed. Testing for synchronicity was carried out on the

‘independent computer’ using Windows 7

and WLM 2011,

by accessing the messages in WLM and the Hotmail

account. This would determine whether the tampered and

fabricated messages persisted on the Hotmail

server when

accessed by the ‘independent computer’.

Outcome 1A: Complete and permanent synchronisation of the tampered

messages was observed. All tampered files were

synchronised and hash values were found to be identical. The

tampered messages persisted in the WLM account and on the

Hotmail

server for a period of eight weeks.

Test 1B: We attempted to replicate the DeltaSync

synchronisation

process on a server to identify the nature of the metadata

synchronised between the server and WLM. This involved

using a server to simulate what occurs on the Hotmail

server

after synchronisation with WLM using Microsoft’s

DeltaSync

, running a similar application, rsync to simulate

DeltaSync

as close as possible, to identify the type of

metadata synchronised.

Outcome 1B: The rsync test showed that unlike Windows, which provides

three distinct types of temporal metadata (created, modified

and accessed), rsync preserves last modification time but

Journal of Digital Forensics, Security and Law, Vol. 7(1)

loses last access time. Modification is a combined creation

and modified time, while accessed is displayed as an access

time. Consequently, server-side metadata and logging is not

expected to provide as much temporal metadata as stored in

the WLM message folders. We recognise there are

applications freely available that are capable of falsifying file

metadata to assist camouflage client-side emails

Question 2: Does WLM support the process of altering email message

content?

Test 2: We added the Hotmail

account to WLM on the ‘editing

computer’. Twelve email messages from the WLM Inbox and

Sent items folders were dragged onto the desktop. Each

message was opened on the desktop with Notepad, the

original message content was deleted and a substituted with a

different message. Using Notepad, the message was then

saved and reinserted into WLM. WLM was synchronised

with the Hotmail

server, closed and the computer rebooted.

Outcome 2: WLM does support the process of altering message content.

The tampered messages persisted in the WLM account and

on the Hotmail

server for a period of eight weeks. These

results were confirmed by accessing the Hotmail

server by

the ‘independent computer’.

Question 3: Does WLM support the process of altering email message

headers?

Test 3: Twelve email messages were dragged to the desktop of the

‘editing computer’ and Notepad was used to view and modify

the message headers. The messages were then reinserted into

their respective folders on the ‘editing computer’. WLM was

closed and the computer rebooted.

Outcome 3: WLM does support the process of altering message headers,

notably message name, message timestamps, IP and SMTP

data. The tampered messages persisted in the WLM account

For example, see:

http://www.techrepublic.com/search?q=http%3A%2F%2Fwww.techre

public.com%2Farticle%2Fbuild-your-skills-learn-to-manipulate-file-

time-stamps-in-windows%2F5034280.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

and on the Hotmail

server for a period of eight weeks. These

results were confirmed by accessing the Hotmail

server by

the ‘independent computer’.

Question 4: Does WLM support the process of altering email message

attachments?

Test 4: During tests 2 and 3, we observed that it was possible to

remove and modify message content script and presumably

possible to remove message attachment script. We wished to

confirm whether it was possible to replace the original

attachment script with script from substitute attachments,

specifically, picture image (.jpg) and Word 2010

document files. To do so, we prepared four similar emails to

the original sent and received messages so that we could

simplify the substitution process by having available

attachment data of the same format as the target message. We

would then transfer that image and text file data by copy and

paste into the original message using Notepad on the ‘editing

computer’. WLM was closed and the computer rebooted.

Outcome 4: WLM does support the process of deleting message

attachments and substituting different attachments; notably

text and image files. The tampered messages persisted in the

WLM account and on the Hotmail

server for a period of

eight weeks. These results were confirmed by accessing the

Hotmail

server by the ‘independent computer’.

Question 5: Does WLM support the insertion of fabricated messages?

Test 5A: Using a template from a different Hotmail

account and

mimicking the header and formatting of existing sent and

received messages, four fresh messages were created and

saved to the ‘editing computer’ desktop. The messages were

then inserted into the Inbox and Sent Items folders of WLM,

which was then synchronised with the server.

Test 5B: We created a fabricated text message in Notepad and saved it

on the desktop as an email message (.eml) to see whether the

fabricated message could be dragged into WLM and

synchronise with the Hotmail

server.

Word 2010

is a registered trademark of Microsoft Corporation.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Outcomes 5A/B: WLM does support the process of inserting fabricated

messages. The tampered messages persisted in the WLM

account and on the Hotmail

server for a period of eight

weeks. These results were confirmed by accessing the

Hotmail

server by the ‘independent computer’.

Identical outcomes were observed using the ‘third computer’ running Windows

with Service Pack 2 and a compatible version of WLM.

6. OTHER OBSERVATIONS

During the course of the study, we recorded observations of additional details

regarding tampering and insertion processes that provide an insight into the

exploit process.

Altering and creating message contents

Using Notepad on the ‘editing computer’ during Test 2 to alter message content

and save the files was a straightforward process only requiring the user to identify

the message content and then substituting the new message by typing or pasting

new text and saving the file. Identification of the message content in the original

message was the first step so that it could be removed or altered, and with a little

practice and intuition, might be considered that a task a novice user could

complete successfully. Figure 5a shows a portion of the message script opened in

Notepad with the message content ‘No attachment’ clearly displayed. Figure 5b

illustrates the message content replaced with the word ‘substituted’. The file is

saved in Notepad and dragged back into the WLM folder. These messages

remained in the same form when viewed on the ‘independent computer’ when

viewed in WLM and the Hotmail

account. All tampered messages appeared

authentic with no evidence of tampering inside the message script.

Messages remaining on the ‘editing computer’ desktop were deleted and the

Recycle Bin emptied. Recovery of erased messages was possible using the

forensics recovery tool and would be of potential evidentiary value in verifying

tampering.

Altering and creating message names, timestamps and other header data

During Test 3, the process of identifying the relevant timestamps within the

extracted messages required some testing on the ‘editing computer’ to ensure that

all relevant data was changed. This was important, as tampering would be evident

if all temporal data, typically three date sets in the header, did not match and the

reinserted file showed the original date and time in the folder view. Initially, some

reinsertion attempts did fail because of difficulties identifying the header data; for

example, Hotmail

headers are different from Gmail, Yahoo etc. Perseverance

and practice rewarded our efforts in identifying critical dates and times ensuring

they were consistent with the original message format. These messages were

Journal of Digital Forensics, Security and Law, Vol. 7(1)

reinserted and did move to the intended new time-based order in the message

folders in accordance with the changed dates and times. The files retained their

position and format throughout the eight-week period.

Figure 5a. Original ‘sent’ message with the string, “No Attachment” visible in

two locations.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Figure 5b. Original message replaced with the tampered ‘sent’ message. The

string “substituted” visible at the same locations. Note the date/time and

message name changes.

Changing the message name was a simpler process to master and it was possible

to change all message names, which retained their status after reinsertion. Figure

6a shows the original message name ‘ORIGINAL.docx’, whereas Figure 6b

shows the altered message name changed to ‘SUBSTITUTED

DOCUMENT.docx’.

Changing IP and SMTP data was a straightforward process although a forger

would have to ensure that the altered data would enhance the forgery. For

example, if a fabricated IP address was inserted this might reveal tampering and

defeat the purpose of the forgery. If an attacker manipulated a message the

metadata stored locally, the Modified, Accessed, Created times will change and

be identifiable to a forensic examiner. Specifically, those metadata would differ

from the header information that is contained in the message when they should be

in close approximation of one another.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

From our observations, we identified and analysed groups of evidence artifacts

created during the tampering processes on the ‘editing computer’. This included

locating evidence of tampering and the use of the WLM exploit on the ‘editing

computer’. From the outset, it was evident that WLM logs, while a possible

source of ‘tamper evidence’ may not always be available for examination if the

forger removed traces of the application’s presence on a computer used to

complete the tampering or fabrication. Similarly, effective erasure of email

messages edited on the host computer desktop would leave little for the forensic

examiner to use.

Figure 6a. Original message showing text attachment script before alteration.

Figure 6b. Tampered message with the document attachment script removed

and replaced with a substitute document script.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

We assume, based on the rsync test and the persistence of messages retained on

the Hotmail

server, that the server would retain creation/modified metadata

when the message was synchronised on the Hotmail

server. Whether the

metadata would provide meaningful corroboration would require checking the

metadata under legislative power or court order. This metadata may be the best

means of authenticating messages timestamps by comparing them with server-

side timestamps. That is, of course, provided the message was still stored on the

server and whether correlating the timestamps was possible to provide meaningful

analysis.

It appears the messages headers scripts are persistent and remain with the message

irrespective of synchronisation with the server. How reliable are the message

header timestamps and can they be taken at face value? Even if the message were

authentic, the timestamps may be inaccurate because of a range of delays and

transmission faults. Comparing the sent message header with the received

message header would provide a means of triangulating the most reliable dispatch

and receipt times. Comparing the messages with the server timestamps is also

conditional on the correct interpretation of the timestamps. Even if they do

conflict with the message timestamps, it may suggest tampering or some other

event. The difficulty is the last accessed time is not overly helpful and the

creation/modified timestamp metadata identified in the rsync test does not provide

a log of original creation or modified events. However, the creation/modified

metadata used to compare the message timestamps would establish message

authenticity.

Deleting and substituting message attachments.

Removing original attachments (Word 2010 documents and image files) from

messages stored on the ‘editing computer’ and replacing them with substitute

attachments initially appeared more challenging and time-consuming than

replacing message and header scripts. Study of the attachment required

confirmation of the start and end of the attachment scripts. Once located the

attachment boundaries were identified it was a straightforward process to remove

original scripts using Notepad. Figure 6a is a segment of an original message

viewed in Notepad showing the start and end boundaries of the attachment script -

a Word 2010

document in this instance. Once removed, it was possible to copy

and paste the substitute attachment script from a previously prepared message

attachment into the tampered message. Figure 6b shows a segment of the

substituted attachment script (Word 2010

) inserted into the message.

Comparison of the hash values of all messages and attachments reinserted into

WLM showed they maintained all their file characteristics.

Ability to alter messages in the directory folder

In a separate test using the ‘editing computer’, we noted that the WLM exploit did

not support attempts to open and alter messages located in the computer directory

Journal of Digital Forensics, Security and Law, Vol. 7(1)

folders that were visible using Windows Explorer

. While fabricated messages

could be inserted and existing messages could be opened using Notepad, the

messages did not synchronise with WLM and did not appear in the folders when

viewed with WLM. None of these messages synchronised with the Hotmail

server because of the local synchronisation that only synchronises messages when

WLM is run.

Ability to create fabricated messages

Test 5 showed that it was possible to create and insert fabricated messages by two

different processes: by creating an email message, and by changing the file

extension of a text document to .eml format. Both processes provide the means to

include malicious code into the message script and pose security threats to unwary

recipients unprotected by malware filters.

7. SUMMARY OF FINDINGS

Our ability to manipulate and insert messages through WLM into the Hotmail

server established proof of concept of the exploit. Study of these processes

showed that the exploit could be used for message tampering and the insertion of

malicious code. Test 1 confirmed that the tampered messages with altered

messages content persisted on the Hotmail

server for the trial period of at least

eight weeks.

Tests 2, 3 and 4 show that WLM permits tampering through a simple ‘drag and

drop’ message migration using a simple text editor. We observed the process

failed repeatedly when attempting to tamper messages in the computer folder

directory housing WLM files. Tampering with message content is a relatively

simple process requiring relatively little knowledge. Changing file headers and

removing and substituting attachments requires more knowledge, but no advanced

computer skills.

Test 5 confirmed the ability to create and insert fabricated messages into WLM

and synchronise those messages with the Hotmail

server through two different

processes. The process of creating a fabricated message mimicking a genuine

email required some practice, but was helped by using message templates

matching message formatting to other email applications and not considered an

insurmountable challenge to novice forgers. The process of changing file

extensions to .eml was uncomplicated but their format was readily recognisable as

fabricated messages. Of security concern is that this process permits the inserting

of ‘zero-day’ code into a target computer.

Time did not permit us to trial similar examinations with Gmail™, Yahoo 7

Mail

and AOLmail applications beyond some initial account creation in WLM

and basic proof of concept trials. Preliminary trialing did confirm that tampering

Windows Explorer

is a registered trademark of Microsoft

Corporation.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

was achieved to the same degree as with our earlier examinations of Hotmail

Outlook 2010

appeared to present a similar opportunity for a similar process to

be exploited.

There may be other applications, such as Mozilla Thunderbird that enables

message tampering with similar results as WLM. We leave that to further research

to clarify the possibility but consider it worthy of mention at this time. Similar

attacks may be feasible against Gmail using IMAP.

8. IMPLICATIONS FOR DIGITAL EVIDENCE ANALYSIS AND

RELIABILITY

Having established the vulnerability of various web-based email applications to

tampering, we can envision serious third-party injection attacks. Consider a

government official with home and office internet access. The attacker injects

malware with appropriate social engineering tags inserted into the message and is

synchronised to the investigator’s account as we have described previously.

Custom code ('zero day') can pass virus scans and is later run by the investigator

as a trusted programme potentially compromising the official’s home and office

network.

In another scenario, a vengeful, disgruntled client seeks to implicate a business

provider and tampers with an existing email from the provider to discredit the

latter. If the provider has no record to contradict the forgery, it may be

burdensome if not impossible to support a counter-claim, for as we have shown,

there may be a paucity of tampering evidence.

A tampered email message could have serious implications for its victim as well

as potential benefits for the forger. We assert that Hotmail

messages tendered as

evidence should not be accepted at face value if there were even the slightest

doubts over their integrity; yet some form of validation is required. Validation

requires confidence about inferences drawn from the evidence, in particular,

verification of the domain where the evidence is created, processed and

transferred (Boddington, Hobbs & Mann, 2008). As Boddington et al., (2008)

assert, not only the validity of the evidence files but examination of the

application and operating programs must be available for examination. Simply

assuming an email message is authentic and relates to a critical date and time,

without seeking some reasonable validation can be disastrous. Such expediency or

inattention to detail, as Dardick (2010) points out, may involve faulty reasoning

that fails to prove that the facts support the conclusions and only those

conclusions. It was important to see whether forgeries and fabricated messages

could be detected though.

A potentially rich source of evidence is the directory folder and WLM but a

perceptive forger may well deep erase traces of WLM, or it may be unavailable if

the ‘editing computer is not located or its existence unknown. We assume, based

on the rsync test and the persistence of messages retained on the Hotmail

server,

Journal of Digital Forensics, Security and Law, Vol. 7(1)

that the server would retain creation/modified metadata when the message was

synchronised on the Hotmail

server. Whether the metadata would provide

meaningful corroboration would require checking the metadata under legislative

power or court order. This metadata may be the best means of authenticating

message timestamps by comparing them with server-side timestamps, which

depends on the message still being stored on the server and whether correlating

the timestamps provides meaningful information to assist analysis.

What concerns us, and should certainly be heeded by those with a vested interest

in seeking the truth, is that while tampering may be suspected, the metadata

available to support or refute assertions as to the message integrity may be scarce

in cases where attempts have been made to camouflage the alterations.

Government and private organisations often have in place email security and pre-

forensics strategies that record all received, sent and deleted messages on their

own servers. This would seem to be a wise precaution, because such strategies do

save deleted messages for use in future investigations. Having an original

message to compare with a tampered message may be a sound validation process.

Otherwise, if an original message is no longer available there is nothing to

challenge a tampered or fabricated message linked to the server with a wrongdoer

exploiting an external Hotmail

account.

Further research would be helpful to understand more fully, what useful server-

side evidence is available to the forensic practitioner. Equally important would be

a reliable, formal validation template to help the forensics practitioner examine

suspect messages. Given the large number of email clients in use, the question of

email aggregation playing a role in this type of exploit should also be

considered. The WLM exploit may be prevented by administration through one of

the other email clients, and since the exploit deals with the messages maintained

on the account holder’s computer and their synchronization to the Hotmail

server, this raise the question of whether this can be bypassed by using another

client. While the main focus of this paper is on WLM and how it stores

messages, these other aspects warrant further consideration.

9. CONCLUSIONS

The results of this study may well have implications for cyber forensics examiners

and legal practitioners preparing cases involving email evidence. We have shown

that WLM facilitates forgery and requires little computer skill to fulfill improper

or mischievous aims. Validation of email messages should be undertaken

whenever there is a suspicion of, or claim of tampering is implicit. This requires

access to email message metadata server side and from likely venues where

WLM was used to undertake tampering.

This confirmation of the WLM process to edit and exploit web-based emails

persuaded us to bring this to the attention of law enforcement authorities. We

hope it will assist forensics examiners in considering the possibility of message

tampering in future cases and perhaps review some existing and previous cases

Journal of Digital Forensics, Security and Law, Vol. 7(1)

where incorrect assumptions were made regarding the authenticity of email-based

digital evidence.

Assuming the server side messages are stored in some sort of database (and a

reasonably complex one we assume), the problem could be fixed by removing

updates to message bodies and attachments within the server side database, so that

when synchronisation occurs, only the entire email can be removed or deleted,

rather than updated. The implications of the WLM exploit process should be

considered by Microsoft and fixed.

Civil, criminal and internal disciplinary cases involving emails are not novel and

occur with increasing sophistication. We predict cases of tampering will come to

public note more often in the future. Pandora’s email box is now opening and

legal cases relying on email evidence should be cognisant of the danger of

assuming the obvious. Prudence suggests circumspection of the circumstantial.

AUTHOR BIOGRAPHIES

Mr. Richard Boddington holds a B.Sc (Hons) 1

Class and is undertaking Ph.D.

research in digital evidence validation at Murdoch University, Australia where he

teaches and researches information security and cyber forensics. He has a police

and security intelligence background and provides cyber forensic analysis and

expert testimony for the legal fraternity in a range of civil and criminal cases.

Mr. Jeremy Ardley has a B.Sc in Physics and Computer Science. He has 30+

years’ experience in information capture, analysis, presentation, and advice to

decision makers. He now works as a Forensic Information Analyst, interpreting

forensic evidence and advises legal teams on technical evidence as well as acting

as an expert witness.

Mr. Grant Boxall is a third year student completing a B.Sc, majoring in Business

Information Systems, Cyber Forensics and Computer Science at Murdoch

University where he is also completing a postgraduate certificate in Business

Administration. He has a background in database programming, data analysis

and application development.

ACKNOWLEDGEMENTS

Our sincerest thanks are extended to Drs. Valerie Hobbs and Graham Mann of

Murdoch University for their support and feedback during the preparation of the

paper.

REFERENCES

Alexander, A. (2008). JT Shannon Lumber Company, Incorporated versus

Gilco Lumber Incorporated. Mississippi: United States District Court, Northern

District of Mississippi, Delta Division.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Ardley, J. (2011). Personal communication with Microsoft: Live Mail

(Hotmail) functionality. (Affidavit). Perth, Australia.

Attorney-General’s Department. (2010). Carrier-carriage service provider

data set consultation paper. Retrieved April 17, 2012, from

http://images.smh.com.au/file/2010/07/23/1710367/Secret-

Document.PDF?rand=1279847709475

Boddington, R. G., Hobbs, V. J., & Mann, G. (2008). Validating digital

evidence for legal argument. Paper presented at the SECAU Security

Conferences: The 6th Australian Digital Forensics Conference, Perth, WA.

Craddock, D. (2010a). A short history of Hotmail. Retrieved April 17, 2012,

from

http://windowsteamblog.com/windows_live/b/windowslive/archive/2010/01/06

/a-short-history-of-Hotmail.aspx

Craddock, D. (2010b). Hotmail now supports push email, calendar, and

contacts with Exchange ActiveSync, Inside Windows Live.

Craddock, D. (2011). Hey! My friend’s account was hacked! Retrieved April

17, 2012, from

http://windowsteamblog.com/windows_live/b/windowslive/archive/2011/07/14

/hey-my-friend-s-account-was-hacked.aspx

Dardick, G. S. (2010). Cyber forensic assurance. Paper presented at the 8th

Australian Digital Forensics Conference.

Florencio, D., & Herley, C. (2007). A large-scale study of web password

habits. In: WWW. Proceedings of the 16th International Conference on World

Wide Web, New York, American Computer Magazine.

LeBlanc, B. (2007a). Introducing Windows Live Mail. Retrieved April 17,

2012, from

http://windowsteamblog.com/windows/b/windowsexperience/archive/2007/05/

07/introducing-windows-live-mail.aspx

LeBlanc, B. (2007b). Microsoft Outlook Connector beta now available.

Retrieved April 17, 2012, from

http://windowsteamblog.com/windows/b/windowsexperience/archive/2007/06/

11/microsoft-office-outlook-connector-beta-now-available.aspx

Levi, A., & Koc, C.K. (2001). Inside risks: Risks in email security.

Communications of the ACM, 44(8): 112.

Preibusch, S., & Bonneau, J. (2010). The Password Game: Negative

Externalities from Weak Password Practices. In Alpcan, Buttyán, and Baras

(Eds.), Decision and Game Theory for Security. Heidelberg, Springer-Verlang.

6442: 192-207.

Journal of Digital Forensics, Security and Law, Vol. 7(1)

Sierra, P. (2010). What draws people to Windows Live Mail and other email

applications? Inside Windows Live.

Sunner, M. (2005). Email security best practice. Network Security, 2005(12):

4-7.