In Re FACEBOOK INTERNET TRACKING LITIGATION

Filing 91

*** POSTED IN ERROR *** please see #93 REDACTED VERSION OF SECOND AMENDED CONSOLIDATED CLASS ACTION COMPLAINT against All Defendants. Filed by Perrin Aikens Davis. (Attachments: #1 Exhibit A, #2 Exhibit B, #3 Exhibit C, #4 Exhibit D, #5 Exhibit E, #6 Exhibit F, #7 Exhibit G, #8 Exhibit H, #9 Exhibit I, #10 Exhibit J, #11 Exhibit K, #12 Exhibit L, #13 Exhibit M, #14 Exhibit N, #15 Exhibit O, #16 Exhibit P, #17 Exhibit Q, #18 Exhibit R, #19 Exhibit S, #20 Exhibit T, #21 Exhibit U, #22 Exhibit V, #23 Exhibit W, #24 Exhibit X, #25 Exhibit Y, #26 Exhibit Z, #27 Exhibit AA, #28 Exhibit BB, #29 Exhibit CC, #30 Exhibit DD, #31 Exhibit EE, #32 Exhibit FF, #33 Exhibit GG, #34 Exhibit HH)(Kiesel, Paul) (Filed on 11/30/2015) Modified on 6/3/2016 (cv, COURT STAFF). Modified on 6/3/2016 (cv, COURT STAFF).

Exhibit Z

Token Attempt: The Misrepresentation of Website Privacy Policies
through the Misuse of P3P Compact Policy Tokens
Pedro Giovanni Leon, Lorrie Faith Cranor, Aleecia M. McDonald, Robert McGuire

September 10, 2010
CMU-CyLab-10-014

CyLab
Carnegie Mellon University
Pittsburgh, PA 15213

Token Attempt: The Misrepresentation of Website Privacy
Policies through the Misuse of P3P Compact Policy Tokens
Pedro Giovanni Leon, Lorrie Faith Cranor, Aleecia M. McDonald, Robert McGuire
Carnegie Mellon University, Pittsburgh, PA

ABSTRACT

tokens [17]. Internet Explorer (IE) makes cookie-filtering decisions by comparing a website’s CPs with user-configured
privacy preferences [9]. If a CP does not reflect the website’s actual privacy practices then that CP is not useful for
decision making. In 2002, regulators from several countries
agreed that a P3P policy is legally binding and “constitutes a
representation to consumers on which they can be expected
to rely” [7].
In this paper, we present the results of our automated
evaluation of P3P CPs collected from 33,139 websites. Verifying the accuracy of CPs requires comparing the computerreadable statements in a CP with a website’s actual privacy
practices with respect to cookies. Sometimes this can be
done by reading the statements that sites make in their privacy policies. However, sometimes privacy policies do not
have sufficient details, and they may not be completely accurate. Even when accurate details are available, reviewing the
accuracy of CPs is a very labor-intensive process. Therefore
we looked for ways to identify errors in P3P CPs that can
be determined based on the syntax errors and conflicting tokens within CPs themselves, without having to review thousands of privacy policies, and without requiring first-hand
knowledge that these policies are accurate. We developed
heuristics to detect three categories of CP errors: invalid
tokens, missing tokens, and conf licting tokens.
We found that nearly 34% of the CPs evaluated in August
2010 have at least one error in these categories, and more
than half of those with errors omit required information. In
addition to syntax and semantic errors, we found that 79% of
CPs lack a corresponding full P3P policy, which is required
for P3P-compliance. Among the 100 most-visited websites1
we found 48 sites with CPs, 21 of which exhibited CP errors
that our automated analysis could detect. In addition, 41 of
these popular sites with CPs did not have corresponding full
P3P policies. The numbers reported in this paper should be
considered a lower bound for the actual number of CPs with
errors, as CPs may contain other types of errors that require
manual comparison with human-readable privacy policies or
other types of analysis to detect.
We analyzed the impact of CP errors on privacy and found
that these errors could mislead users by misrepresenting privacy practices with respect to cookies. We also determined
the implications of faulty CPs for user agent behavior. We
analyzed practices that appear to be deliberately designed
to bypass IE default privacy filters and found that more
than 97% of incorrect CPs would bypass these default fil-

Platform for Privacy Preferences (P3P) compact policies
(CPs) are a collection of three-character and four-character
tokens that summarize a website’s privacy policy pertaining to cookies. User agents, including Microsoft’s Internet
Explorer (IE) web browser, use CPs to evaluate websites’
data collection practices and allow, reject, or modify cookies based on sites’ privacy practices. CPs can provide a
technical means to enforce users’ privacy preferences if CPs
accurately reflect websites’ practices. Through automated
analysis we can identify CPs that are erroneous due to syntax errors or semantic conflicts. We collected CPs from
33,139 websites and detected errors in 11,176 of them, including 134 TRUSTe-certified websites and 21 of the top 100
most-visited sites. Our work identifies potentially misleading practices by web administrators, as well as common accidental mistakes. We found thousands of sites using identical invalid CPs that had been recommended as workarounds
for IE cookie blocking. Other sites had CPs with typos in
their tokens, or other errors. 98% of invalid CPs resulted in
cookies remaining unblocked by IE under it’s default cookie
settings. It appears that large numbers of websites that use
CPs are misrepresenting their privacy practices, thus misleading users and rendering privacy protection tools ineffective. Unless regulators use their authority to take action
against companies that provide erroneous machine-readable
policies, users will be unable to rely on these policies.

1.

INTRODUCTION

The Platform for Privacy Preferences (P3P) is a World
Wide Web Consortium (W3C) recommendation for specifying website privacy policies in a machine readable format.
Developed as part of an industry self-regulatory effort and
published in 2002, it provides two privacy policy formats:
full policies and compact policies (CPs). P3P full policies
are XML files that represent website privacy policies in detail. P3P CPs summarize website privacy policies regarding
cookies using a string of three-character and four-character

1
Most-visited websites as of August 2010, according to
http://www.quantcast.com/top-sites-1.

This is an extended version of a paper to be presented at the ACM Workshop
on Privacy in the Electronic Society (WPES 2010), October 2010.

1

2.2

ters. Our work identifies potentially misleading practices by
web administrators, as well as common accidental mistakes.
This paper is organized as follows. In Section 2, we discuss
the details of P3P and compact policies and review related
work. In Section 3, we describe our study methodology. In
Section 4, we introduce and define common P3P CP errors,
present data on the frequency with which each type of error
appears, and discuss the privacy implications of these errors.
In Section 5, we analyze the impact of incorrect CPs on IE
and discuss evidence that incorrect CPs are being used to
bypass IE privacy filters. In Section 6, we analyze the CPs
of three groups of websites: popular sites, network advertisers that offer opt-out cookies, and TRUSTe-certified sites.
Finally, we present our conclusions in Section 7. We also
include seven appendixes that provide detailed information
about CP tokens and data from our analysis.

2.

Compact policies (CPs) are defined in the P3P specification as an optional performance optimization. CPs are optionally served when a website transmits a cookie. They provide a lightweight mechanism to provide information about
a website’s privacy practices with respect to cookies and help
user agents quickly decide how to process cookies. The P3P
specification requires that sites that choose to deploy CPs
make an effort to do so accurately. According to the P3P
specification, “if a web site makes compact policy statements
it MUST make these statements in good faith” [18]. Despite
being an optional component of P3P, CPs are important because they are used by IE 6, 7, and 8 to determine whether
to block or modify cookies.
P3P specifies a set of CP tokens associated with nine
types of P3P policy elements. Valid CPs must have at least
five of these elements. The valid tokens for each element,
the corresponding full-P3P policy elements, and a short description of each are provided in Appendix A.

BACKGROUND AND RELATED WORK

In this section we provide an overview of the Platform for
Privacy Preferences (P3P) 1.0. In addition, we provide background on P3P compact policies and discuss related work.

2.1

2.3

Internet Explorer Cookie Blocking

IE 6, 7, and 8 allow users to set their privacy preferences,
which are then used to evaluate websites’ CPs and perform
cookie filtering. IE considers cookies that are used to collect
personally identifiable information (PII) without providing
users the choice to opt-in/opt-out to be unsatisfactory cookies. IE performs cookie filtering based on six privacy levels
that the user can set up; namely, Block All Cookies, High,
Medium High, Medium (default level), Low, and Accept All
Cookies [9]. The vast majority of users do not change the
default privacy settings, so analyzing filtering conditions at
the medium level allows us to determine the privacy impact
of invalid CPs that bypass IE filters. Under the medium
setting, unsatisfactory first-party cookies are converted into
session cookies and unsatisfactory third-party cookies are rejected. In addition, third-party cookies not accompanied by
CPs are rejected. When cookies are rejected, website functionality may be impaired and sites may be unable to collect
some of the data they use for operational or business purposes. Because of this cookie-blocking feature, many website
administrators have implemented CPs on their websites.

The P3P Specification

P3P is a standard for specifying websites’ privacy policies in a machine-readable XML format that can be processed and acted upon by automated tools [2], [17]. P3P
allows user agents to automatically evaluate privacy policies against users’ preferences. For example, users can set
up their privacy preferences using user agents embedded in
their web browsers so that their browsers will warn about
mismatches with privacy preferences or block cookies at sites
where mismatches occur [4]. When P3P is embedded in
search engines, search results can be annotated automatically with privacy information [3]. P3P enables automatic
generation of standardized “nutrition label” privacy notices,
which are more understandable and easier to read than conventional policies [8]. It also allows automated tools to collect privacy policy data for analysis [1].
Published in 2002, the P3P 1.0 specification defines two
types of P3P policies: full policies and compact policies.
P3P full policies are written in XML format and use a defined vocabulary and a data schema to represent humanreadable privacy policies in a machine-readable language.
Compact policies summarize privacy practices associated with
the use of cookies; they are transmitted in response to HTTP
requests using HTTP headers. Full P3P policies are the authoritative source for website information management practices. The P3P specification requires compact policies to
have corresponding full P3P policy files [17].
The P3P specification defines a protocol for requesting
and transmitting P3P policies via HTTP. Retrieving a P3P
policy requires a two-step process. P3P user agents issue requests for P3P policy reference files stored in a “well-known”
location and also check for P3P HTTP headers and P3P
metadata embedded in HTML content. P3P policy reference files contain references to the location of full P3P policy
files. After fetching a P3P policy reference file, user agents
are able to locate and retrieve a full P3P policy file [17].
W3C maintains an online validation tool that checks for
syntax errors in full P3P policies and CPs.2
2

P3P Compact Policies

2.4

Related Work

The first large-scale automated analysis of P3P policies
found that approximately 10% of 5,739 websites tested in
2003 were P3P-enabled [1]. In addition, 85 websites had only
compact policies and were therefore not P3P-compliant, and
about a third of the full P3P policies had technical errors.
Reay et. al. performed a study of over 3,000 full and compact P3P policies. They found high rates of websites posting CPs without corresponding full P3P policies, as well as
many sites that had conflicts between full and compact P3P
policies [14].
Cranor, et al. performed an error analysis on P3P full
policies. They found that 73% of the 14,720 full P3P policies analyzed contained syntax errors. They compared full
P3P policies with their corresponding natural language privacy policies at 21 popular websites and found conflicts between the P3P and natural language policy at each of these
sites. Most conflicts were associated with the P U RP OSE,
CAT EGORIES, and RECIP IEN T S elements [3].
Reidenberg and Cranor studied the accuracy of P3P user

http://www.w3.org/P3P/validator.html
2

agents [15]. They suggested that inaccurate representations
by user agents could undermine the purpose of the P3P standard. From a legal standpoint, the enforceability of an agreement based on a P3P user agent’s simplified representation
of a website privacy policy is uncertain [6]. In the United
States, the Federal Trade Commission Act of 2006 empowers the FTC to prevent “unfair or deceptive acts or practices
in or affecting commerce....” The FTC has used this authority to take action against companies with deceptive privacy
policies. While the FTC has not taken such actions on the
basis of deceptive machine-readable privacy policies to date,
it appears to be within the FTC’s authority to do so [7].

3.

three categories: invalid tokens, missing tokens, and conf licting tokens. We define each type of error in Section 4.
We developed scripts to parse CPs and detect each type of
error we defined. We also developed a script to check each
CP to determine whether it would be considered satisfactory
by IE under the default privacy setting [9].

4.

COMPACT POLICY ERRORS

In this section we present the results of our evaluation
of 33,139 collected CPs. We define three categories of errors, and for each type provide data on the frequency of
occurrence. We suggest probable underlying causes of these
errors: typos, lack of understanding of the P3P specification, or attempts to avoid web browser cookie filtering. We
further evaluate the impact of these errors on users’ privacy.
We found 11,176 CPs with errors, about 34% of the URLs
we analyzed. If we aggregate these invalid CPs by unique
domain names, the number of incorrect CPs is reduced to
4,696. Domain names with more than one website account
for at least 57% of the total invalid CPs. If companies hosting multiple sites correct their CPs, the CPs with errors
could be reduced to less than 15% of the URLs we analyzed.
Table 1 summarizes the CP errors we found.
According to the P3P specification, CPs must be accompanied by a full P3P policy. Therefore, we investigated how
many of the collected CPs have a corresponding full P3P
policy. We found that only 21% of those websites providing
CPs also provided full P3P policies.

METHODOLOGY

We collected CPs from four data sets. First, we collected 52,156 URLs containing CPs from the Privacy Finder
search engine cache in January 2010. The cache was initially seeded in 2005 through a web crawl [3] and has grown
over time as a result of Privacy Finder users’ searches. Second, on July 3, 2010 we collected a list of the 3,417 websites
holding TRUSTe privacy seals, as reported by the membership list posted on the TRUSTe website [16]. Third,
we collected a list of the 100 most-visited websites from
www.quantcast.com on August 10, 2010. Fourth, we collected a list of 75 network advertisers offering opt-out cookies that can be set using the Beef Taco (Targeted Advertising
Cookie Opt-Out) Firefox extension as of June 23, 2010. We
removed duplicate domains from multiple datasets, so we
had a final list of 55,636 unique URLs.
In August 2010 we used our web crawler to visit our list of
55,636 URLs and attempt to download CPs. Some datasets
had URLs that were not fully qualified (for example, foo.com
instead of http://www.foo.com) so we prepended http://
and http://www. and tried again if sites failed. When we
contacted sites that no longer responded, we retried twice.
When we contacted sites that gave other errors, we retried
up to ten times. Some of the sites on this list were no longer
available, and some that previously had provided CPs were
no longer providing them. We were able to collect CPs from
33,139 sites.
At sites with P3P CPs we also checked for full P3P policies. The crawler checked for a P3P policy reference file in
the P3P well-known location, HTTP header, and page content. If found we parsed this file to retrieve the location of
the full P3P policy (the policy is either embedded in the
policy reference file or referenced by a URL in the policy
reference file). We then retrieved the full P3P policy and
file and verified that it contained a P3P policy. 7,016 of the
sites with CPs also had full P3P policies.
Some domains we analyzed had more than one host on the
domain with a compact policy in our Privacy Finder cache
dataset, for example both http://www.x.example.com and
http://www.y.example.com. We report data in both aggregated form where each domain is represented only once
for each unique CP found regardless of the number of hosts
(just one entry for example.com if all hosts in example.com
have the same CP) and in non-aggregated form where each
host is represented (x.example.com and y.example.com are
separate). Furthermore, if a single domain exhibited different incorrect CPs, we reported them separately and count
both of them in the aggregated data set. There are 19, 820
domains in our aggregated data set.
We define a set of P3P CP errors and group them into

4.1
4.1.1

Invalid Tokens
Definition

All valid CP tokens are shown in Appendix A. Some
tokens can optionally be accompanied by an attribute that
specifies that a data practice is performed always, on an
opt-in basis, or on an opt-out basis (a, i, or o). No other
tokens or token-attribute combinations are valid. An invalid
token error occurs when a CP includes an invalid token or
an invalid token-attribute combination.
Tokens that do not specify the optional attribute default
to the always attribute (a). The CU R (current purpose)
token does not take an optional attribute, but we found
CU Ra is commonly used in CPs. As CU R is the only
P U RP OSE element that does not allow an attribute, we
believe web administrators may be mistakenly adding the
invalid a attribute. Since CUR already means that data is
always collected for the current purpose this does not change
the semantics of the CP, though it is technically invalid. In
this paper we report the frequency of CU Ra separately from
other invalid tokens and do not count this error in our overall
error numbers.

4.1.2

Evaluation

Of the total CPs evaluated, 11.6% contained invalid tokens. An additional 16% contained the invalid and harmless
CURa token and no other invalid tokens, but we do not include these URLs in our count of sites with invalid tokens.
Companies publishing multiple websites under a single domain name account for 79% of the invalid token errors.

4.1.3

Analysis

Many invalid token errors are likely to be accidental. We
found that several CPs include the CU S string which is not
3

Type of error

Number
of URLs

Percent of
URLs with Errors

Number
of Domains

Percent of
Domains with Errors

Any problem

11,176

33.7%

4,696

23.7%

Invalid tokens
Invalid tokens
CURa (no other problems)

3,839
5,295

11.6%
16.0%

794
2,557

4.0%
12.907%

Missing tokens

6,402

19.3%

3,319

16.8%

Conflict between tokens
IVA
CON
IVD
TEL
NID
NOR
NON

3,977
3,899
1,059
340
366
345
32

12.0%
11.8%
3.2%
1.0%
1.1%
1.0%
0.1%

923
835
669
178
168
99
27

4.7%
4.2%
3.4%
0.9%
0.9%
0.5%
0.1%

Total analyzed

33,139 URLs

19,820 Domains

Table 1: Summary of CP errors. Some CPs have errors in more than one category. CPs that contain the
invalid CU Ra token and no other errors are not included in the “any problem” count.
a valid token. We believe that web administrators might
have intended to use the T AI (tailoring) token but wrote
CU S (customize) instead, which is an option that appeared
in early drafts of the P3P specification. Another common
syntax error is swapping letters within tokens. For example, web administrators tend to write OP T instead of OT P ,
T IA instead of T AI and so on. Web administrators commonly add attributes to tokens that do not accept them. For
example, they use CAOo instead of CAO (providing user access to contact information). Another common token found
with invalid attributes was OU R, which is the only recipient
token that does not accept an attribute. Some invalid CPs
contain multiple valid tokens concatenated together without
the required white-space separator. More concerning are the
CPs that contain completely invalid strings such as AM ZN
or V P RT .
While some of the invalid token errors are likely typos
and many appear to be harmless, these errors may cause
user agents to incorrectly interpret a CP, which could confuse and mislead users. In addition, it appears that some
of the completely-invalid tokens are being used by sites to
prevent their cookies from being blocked without providing
a meaningful CP.

4.2
4.2.1

mine a website’s data practices with respect to cookies. For
example, a CP that contains no CATEGORIES token fails
to communicate the type of information associated with a
cookie. Because P3P declarations are positive, the absence
of a token is an indicator that a website does not engage in a
particular practice. Therefore, sites that are missing CATEGORIES tokens are effectively stating that they do not collect any category of data. Furthermore, IE analyzes CATEGORIES, PURPOSE and RECIPIENT tokens to make
cookie-blocking decisions. As we discuss later, if these tokens are omitted, IE may incorrectly accept a cookie that
would otherwise be blocked.

4.3
4.3.1

Missing Tokens
Definition

Evaluation

Missing-token errors are the most common type of error
we found. These errors occur in 19% of the total analyzed
CPs and account for more than 57% of the incorrect CPs.

4.2.3

Definition

The CP semantics should guarantee consistency between
different tokens. For example if a CP includes the ALL token, indicating users have access to all of their personally
identifiable information, the CP cannot include the N ON
token as well, as that means users have no access to their
personally identifiable information. Similarly, if a CP includes the N OR token, which means the website does not
store permanently the information collected, it cannot include any other RETENTION element tokens. If a website
declares that it does not collect personally identifiable information (PII) then the CP should not include tokens that
suggest collection of such information. Several conflicting
token scenarios are defined in the following paragraphs.
Conflict with the NID token. The non-identifiable token
N ID should be used only when the website does not collect
any type of PII and associate it with a cookie. There are
many tokens that suggest collection of this information. In
particular P HY (information to locate or contact an individual in the physical world), ON L (information to locate
or contact an individual on the Internet, e.g. email address),
F IN (financial information), LOC (physical location data)
and GOV (government identifier, e.g. social security number) tokens are directly associated with the collection of PII.
Similarly, CAO (contact and other information) and IDC
(online and physical contact information) ACCESS tokens
should appear only if PII is collected by the website and

According to the P3P specification, each statement in a
CP that does not include the N ID (no user-identifiable data
collected) token should include at least one PURPOSE, RECIPIENT, RETENTION, CATEGORIES and ACCESS token. A missing tokens error occurs when a CP without a
NID token does not contain at least one token from each of
the above five categories.

4.2.2

Conflicting Tokens

Analysis

Missing tokens make it impossible to use the CP to deter4

associated with a cookie. Furthermore, the IV A (individual analysis), IV D (individual decision), CON (contact)
and T EL (telemarketing) PURPOSE tokens require PII and
should not be in the CP if the N ID token is also in the CP.
Conflict with the IVA/IVD tokens. Performing individual analysis (IVA) or making individual decisions (IVD) requires identifying a particular individual. The IVA and IVD
tokens must be accompanied by at least one of the following
CATEGORIES tokens: PHY, ONL, FIN, PUR, GOV.
Conflict with the CON token. The contact token, CON ,
requires enough information to contact the individual either
by electronic or physical means. CON must be accompanied
by at least a P HY or ON L token.
Conflict with the TEL token. Telemarketing requires phone
numbers, which are part of the physical category. Therefore,
the T EL token must be accompanied by a P HY token.
Conflict with the NON/NOR tokens. The N ON token
indicates users are not allowed to access any data collected
about them. None of the remaining tokens in the ACCESS
element should appear in the CP with the N ON token.
Similarly, the N OR RETENTION token means the website does not store permanently any kind of PII. If there is a
N OR token the CP should not contain any other RETENTION tokens.

4.3.2

Type of error

Full P3P
Policies

Ratio
CPs/Full

None (valid CPs)

21,963

5,915

26.9%

Any problem

11,176

1,106

9.9%

Invalid tokens

3,839

255

6.7%

Missing tokens

6,402

469

7.03%

Conflicting tokens
IVA
CON
IVD
TEL
NID
NOR
NON

3,977
3,899
1,059
340
366
345
32

586
321
424
107
127
64
1

14.7%
8.2%
40.0%
31.5%
34.7%
18.6%
3.1%

Table 2: Full P3P policies at websites with CPs.
egories can also lead to user confusion. These errors suggest
either a lack of understanding of the P3P tokens or an attempt to misrepresent a site’s practices.
Most of the websites in the NON conflicting token error
category simultaneously include the CAO and N ON tokens
in their CPs. This suggests a possible misunderstanding of
the N ON token. It leaves users with uncertainty about the
access options offered by the website.

Evaluation

The most common errors in this category are associated
with the individual analysis (IVA) token. We found these
errors in 12% of the analyzed CPs and 35% of the invalid
CPs. The second most common type of error in this category is associated with the CON token, which represents
more than 11% of the collected CPs and more than 34% of
the total incorrect CPs. Conflicts involving the remaining
tokens are relatively rare, each occurring in less than 4% of
the CPs we analyzed.

4.3.3

Number
of CPs

4.4

CPs Without Full P3P Policies

The P3P specification requires websites implementing CPs
to have a corresponding full P3P policy. Only 7,016 of the
33,139 URLs in our data set (21%) had full P3P policies.
As shown in Table 2, CPs without errors were almost three
times more likely to have full P3P policies than those with
errors. 26.9% of error-free CPs had corresponding full P3P
policies, while only 9.9% of CPs with errors had full P3P
policies. For CPs with the two most common errors, invalid tokens and missing tokens, the percentage of URLs
that present full P3P policies is even lower. This finding
provides some evidence that websites with these types of errors may be providing inaccurate CPs to avoid having their
cookies blocked.
We found full P3P policies for 17% of the 391 TRUSTe
sites with CPs, 55% of the 11 network advertising sites with
CPs, and 15% of the 48 most-visited sites with CPs. Appendices E, F, and G show the details on most-visited sites,
network advertiser sites, and TRUSTe sites respectively.
The CPs with IV D, T EL, and N ID conflicting tokens
were more likely to have corresponding full P3P policies.
This result suggests that many of these websites may be
making good faith efforts to properly comply with the P3P
specification and not just sending CPs to avoid cookie blocking. However, web administrators may not fully understand
the meanings of these tokens.

Analysis

When a website incorrectly uses the N ID token, users
have conflicting information about whether or not their PII
is being collected. Web administrators may misunderstand
the definition of the N ID token in the P3P specification and
use it incorrectly. The most common N ID conflict we found
was with the ON L token (email address or other online contact information). This suggests that web administrators
might be unaware that email addresses are considered identifiable information. Another common conflict is with the
IVA token (individual analysis) which implies that identifiable information is used to perform an analysis. If the N ID
token is used and it is true that no PII is collected, the P SA
token (pseudo-analysis) should be used instead of the IV A
token.
The NOR conflicting token error leads to confusion about
website retention practices. This is probably an accidental
mistake, but could negatively affect users’ decisions. We
found that some CPs include N OR and IN D tokens: while
the company claims no retention of PII at the same time
they claim that they store information indefinitely. Many of
the invalid CPs in this error category include the BU S token, meaning the retention period is based on their business
practices. While it might be the case that their business
practices do not require the retention of personal information, including both of these tokens is confusing.
The IVA, IVD, CON and TEL conflicting token error cat-

4.5

Other Semantic Errors

In addition to the errors already discussed, CPs may contain other types of errors that we were unable to detect
automatically. For example, CPs may be semantically inconsistent with human-readable policies posted on websites
or with full P3P policies. In addition, CPs may fail to accurately represent website data practices. Evaluating semantic
errors requires analyzing websites’ human readable policies,
which is not a task that can be automated. Given the large
5

scale of our study, reading human readable policies for the
thousands of sites we evaluated is impractical. Furthermore,
human-readable policies do not always contain detailed information about data collection and treatment associated
with the use of cookies, so even reading these policies would
not necessarily resolve questions of P3P accuracy. Thus, a
complete evaluation of semantic errors is beyond the scope
of our study. However, to gain some insights into the prevalence of semantic errors, we manually compared CPs with
full P3P policies and human-readable policies at 41 sites, including 11 most-visited web sites. As detailed in Appendix
C and D, 40 of these sites failed to provide full P3P policies,
15 failed to provide human-readable privacy policies, and
2 sites had CPs consisting only of meaningless, invalid tokens. When comparing the full policies and human-readable
policies with their corresponding CPs, we found 4 sites with
slight discrepancies, and 15 sites with major discrepancies.
Furthermore, we were not able to compare 5 sites because
their human-readable policies did not include any statement
about cookies.

5.

or CAT EGORIES tokens. Even if the CP were valid, Microsoft’s recommendation undermines the purpose of P3P
since it encourages web administrators to use CPs that do
not represent their actual data practices. We found several
technical blogs recommending similar solutions [11], [19].
Some of them suggested the exact CP described above and
referred to the Microsoft support website as the source of
their advice [12]. This CP occurred 2,756 times in our data
set. Only 31 of these CPs had corresponding full P3P policies. Nearly 25% of the invalid CPs used these tokens, representing 43% of invalid CPs in the missing-tokens error category. We did not find this CP at any of the 100 most-visited
websites.
In an article titled “P3P in IE6: Frustrating Failure” posted
in the O’Reilly blog, the author suggests another “trick” to
bypass IE6 privacy filters. He recommends adding a P3P
CP header that “enables your cookie to survive any privacy
setting” [5]. The CP suggested is: N OI ADM DEV P SAi
COM N AV OU R OT Ro ST P IN D DEM . This CP does
not contain any of the errors we tested for. However, if
web administrators blindly post it without confirming that
it matches their site’s actual practices, they will mislead
users. This CP occurred 4,360 times in our data, representing 13% of the CPs analyzed. Only 12 of these CPs had
corresponding full P3P policies. We did not find this CP at
any of the 100 most-visited websites.
We manually analyzed 30 privacy policies of websites that
use the CP recommended by the O’Reilly blog. 14 of these
30 websites were randomly taken from the domains that
present the most URLs with this CP. The remaining 16
evaluated websites were randomly chosen from the set of
websites sending this CP. From the 30 manually-analyzed
websites, only one had a valid P3P full policy. However,
the valid full P3P policy did not match the corresponding
CP, and when we returned to the website in September we
discovered the full P3P policy had been removed. We were
unable to locate any human-readable privacy policy at 15 of
these websites. Furthermore, none of the human-readable
privacy policies we found properly matched the corresponding CP. Apendix C summarizes the results of this manual
evaluation.
From the 15 websites that exhibited a human-readable
policy, only 10 mention the use of cookies. However, we
found semantic inconsistencies between all of these policies
and their corresponding CPs. Most of these policies made
no mention of data practices related to the N OI access token; ADM , DEV , or P SAi purpose tokens; ST P or IN D
retention tokens; or DEM categories token. Many of them
made references to other cookie-related data practices not
captured by the CP.
We performed a Google search for “ie blocking iframe
cookies” and found a number of sites suggesting similar solutions. For example, another blog post recommended a
CP that we found 300 times in our data set [19]. On the
other hand, we also found a popular question-and-answer
site that advised web administrators to create CPs that accurately reflect their site’s privacy policy: “The tags are not
only a bunch of bits, they have real world meanings, and
their use gives you real world responsibilities! For example, pretending that you never collect user data might make
the browser happy, but if you actually collect user data, the
P3P is conflicting with reality. Plain and simple, you are
purposefully lying to your users....” Immediately under the

COMPACT POLICIES AND INTERNET
EXPLORER

Microsoft helped drive P3P adoption by using P3P CPs to
make cookie-blocking decisions. However, the large number
of CP errors and low rate of full P3P compliance suggest
that many websites are adopting P3P CPs in order to avoid
cookie blocking, but are not presenting accurate representations of their privacy practices. To gain additional insights
into CP adoption, we analyzed the CPs we collected to determine whether IE would classify them as satisfactory under
the default medium setting. We found that 99% of CPs collected would be considered satisfactory by IE. Of the 33,139
CPs we examined, only 118 error-free CPs and 263 CPs with
errors were considered unsatisfactory, and thus would likely
be blocked when the cookie was used in a third-party context. All but three of the unsatisfactory CPs with errors
had missing-tokens errors.
IE cookie filters only look for combinations of tokens considered unsatisfactory. They ignore invalid tokens and do
not check to make sure the minimum required tokens are
present. They also do not look for token conflicts. If IE performed the same sort of checks we did in our analysis and
treated CPs with these errors as unsatisfactory, we would expect the error rate to be reduce over time because companies
would have an incentive to correct the errors in their CPs
(although it is possible that we would then see an increased
rate of other types of errors that are less-easily detectable
through automated analysis).
After observing a large number of identical CPs in our
data set, we suspected that web administrators might be
copying these CPs from a common source. We used a search
engine to track down the source of some of the most common
CPs in our data set.
We discovered that Microsoft’s support website recommends the use of invalid CPs as a work-around for a problem in IE. Specifically, a FRAMESET or parent window
that references another site inside a FRAME considers the
referenced site as a third-party, even if it is first-party content located on the same server [10]. Microsoft suggests
the following invalid CP: CAO P SA OU R. This CP is
clearly invalid since it does not contain any RET EN T ION
6

Compact Policy
listed in http-stats.com/P3P

Number of
occurrences
in collected
CPs

Errors

IE Satisfactory
cookie?

CAO DSP COR CUR ADM DEV TAI PSA PSD
IVAi IVDi CONi TELo OTPi OUR DELi SAMi
OTRi UNRi PUBi IND PHY ONL UNI PUR FIN
COM NAV INT DEM CNT STA POL HEA PRE
GOV

2,738

None

YES

NOI ADM DEV PSAi COM NAV OUR OTRo STP
IND DEM

4,360

None

YES

Missing tokens

YES

NON DSP CURa OUR NOR UNI

0

ALL CURa ADMa DEVa TAIa OUR BUS IND PHY
ONL UNI PUR FIN COM NAV INT DEM CNT
STA POL HEA PRE LOC OTC

102

None

YES

BUS CUR CONo FIN IVDo ONL OUR PHY SAMo
TELo

293

Missing tokens

YES

CAO DSP COR CURa ADMa DEVa OUR IND PHY
ONL UNI COM NAV INT DEM PRE

553

None

YES

NOI NID ADMa OUR IND UNI COM NAV

464

None

YES

ALL DSP COR CURa ADMa DEVa TAIa PSAa PSDa
IVAa IVDa CONa TELa OUR STP UNI NAV STA
PRE

24

Conflicting IVA, IVD,
CON, and TEL tokens

YES

359

None

YES

Missing tokens

YES

NOI DEVa TAIa OUR BUS UNI STA
CAO PSA OUR

2,756

Table 3: Frequency of “common” CPs from http-stats.com in collected CPs
example CP was the warning: “Note that the combination
of P3P headers in the example may not be applicable on
your specific website; your P3P headers MUST truthfully
represent your own privacy policy!” [13]. This warning must
have been effective as we did not find the example CP in our
data set.
We examined the top 10 P3P header values listed at httpstats.com, a website that crawls the web and compiles data
on HTTP header values. These headers included the CPs
recommended in the O’Reilly blog and on the Microsoft support website. As shown in Table 3, we found multiple instances of nine of these CPs in our data set and detected
errors in four of these CPs. All of these CPs are considered
satisfactory by IE.

6.

displayed a full P3P policy.
The f acebook.com CP is invalid because it is missing
PURPOSE, RECIPIENT, RETENTION, CATEGORIES
and ACCESS tokens. The included tokens, DSP and LAW ,
do not provide any information about the site’s collection or
use of data. This CP simply states that any privacy dispute
will be resolved according to a law referenced in their privacy policy, and implies that the site collects no data associated with the cookie. When doing preliminary work for this
study in 2009, the f acebook.com compact policy contained
only the single invalid token HON K. Both of these CPs are
useless for communicating with user agents and users. It is
likely that f acebook.com is using their CP to avoid being
blocked by IE.
Websites under the msn.com domain exhibited a CP that
includes the invalid CU So token. Two other Microsoftowned sites, microsof t.com and windows.com use the same
CP. These websites display the TRUSTe EU Safe Harbor
Privacy seal. We believe that these websites are likely attempting to comply with P3P; however, they are not using
P3P properly.
The live.com CP does not include any ACCESS tokens.
This CP suggests collection of PII, but does not provide any
information about whether users can access their personal
information.
The amazon.com and imdb.com domains each contain a
single invalid token and no other tokens, so they fall into
the invalid-tokens and missing-tokens categories. It appears
that these two websites use a CP only for the purpose of
avoiding IE cookie filtering.
The aol.com domain’s CP is invalid since it is missing
ACCESS and RETENTION tokens. The mapquest.com
domain is missing a RETENTION token. The godaddy.com

COMPACT POLICIES AND POPULAR
WEBSITES

In this section we present the results of our evaluation of
the top-100 most-visited websites, 3,417 TRUSTe-certified
websites, and 75 network advertising websites offering optout cookies that can be set using the Beef Taco Firefox extension. We found that only 391 of the evaluated TRUSTe
sites had CPs and 134 of those had errors. 48 of the 100
most-visited sites had CPs and 21 had errors. 11 of the
evaluated network advertising sites had CPs and only one
had an error.
The top-visited domains with CPs are listed in Appendix
E. We analyzed the errors in detail for the top 50 mostvisited sites with CPs that contained errors. Because of the
popularity of these sites, errors in their CPs have an impact
on a large number of users. Table 4 shows the CPs and types
of errors found. Only one of these websites, microsoft.com,
7

URL

Compact Policy

Errors found

Valid Full
P3P Policy

IE Satisfactory
cookie?

TRUSTe
seal

facebook.com

DSP LAW

Missing tokens

NO

Yes

EU Safe Harbor

msn.com

BUS CUR CONo FIN IVDo ONL OUR
PHY SAMo TELo

Missing tokens

NO

Yes

EU Safe Harbor

live.com

BUS CUR CONo FIN IVDo ONL OUR
PHY SAMo TELo

Missing tokens

NO

Yes

EU Safe Harbor

amazon.com

AM ZN

Invalid tokens,
Missing tokens

NO

Yes

None

microsoft.com

ALL IND DSP COR ADM CONo CUR
CU So IVAo IVDo PSA PSD TAI TELo
OUR SAMo CNT COM INT NAV ONL
PHY PRE PUR UNI

Invalid tokens

YES

Yes

EU Safe Harbor

reference.aol.com

UNI INT STA NAV DEV CUR OUR

Missing tokens

NO

Yes

Web Privacy

atlas.mapquest.comSTA INT UNI CUR DEV NOI OUR

Missing tokens

NO

Yes

None

godaddy.com

IDC DSP COR LAW CUR ADM DEV
TAI PSA PSD IVA IVD HIS OUR SAM
PUB LEG UNI COM NAV STA

Conflicting IVA
and IVD tokens

NO

Yes

EU Safe Harbor

imdb.com

IM DB

Invalid tokens,
Missing tokens

NO

Yes

None

windows.com

ALL IND DSP COR ADM CONo CUR
CU So IVAo IVDo PSA PSD TAI TELo
OUR SAMo CNT COM INT NAV ONL
PHY PRE PUR UNI

Invalid tokens

NO

Yes

None

hulu.com

NOI DSP COR NID ADMa OP T a OUR
NOR

Invalid tokens

NO

Yes

None

Table 4: CPs of 50 most-visited websites with errors, sorted by popularity. Invalid tokens are shown in italics.
domain has conflicting IV A and IV D tokens. This CP is
confusing since it claims the site is using identifiable information to perform individual analysis and decisions; however, it
does not include any CATEGORIES token associated with
the collection of identifiable information.
Finally, the hulu.com domain contains the invalid OP T a
token which presumably is intended to be OT P a instead.
This makes it syntactically incorrect. More importantly, the
CP includes the N ID token, claiming that no PII is associated with cookies. We read the human-readable privacy
policy of this website and found that it explicitly mentions
linking PII to cookies.
All but one of these top websites do not have a full P3P
policy, and several of them have CPs that appear to be
well-crafted to bypass IE filtering. Further analysis will be
needed to determine if they actually follow the data practices they claim through their CPs; however, as detailed in
Appendix D, there are inconsistencies that suggest they do
not follow the practices they claim.
Domains such as f acebook.com, msn.com, live.com and
aol.com exhibited TRUSTe privacy seals, despite displaying invalid CPs. Indeed, we found that 391 of the 3,417
TRUSTe-certified websites have CPs, but 134 (34.3%) of
these had at least one problem with their CPs, as detailed
in Appendix G. 28 out of the 48 top websites with CPs
appeared on the list of TRUSTe websites and 11 (39.3%) of
these had invalid CPs. This suggests TRUSTe is not reviewing websites’ CPs when issuing privacy seals.
Network advertisers tend to make heavy use of third-party
cookies in order to provide targeted advertising. Therefore,
the use of CPs among network advertisers is of particular im-

portance. Without CPs, many network advertising cookies
would be blocked by IE because they are used in a thirdparty context. In addition, users are generally not aware
of what third-parties are setting cookies on the sites they
visit or what their privacy practices are. If used properly,
P3P could provide information about privacy practices that
would otherwise be difficult for users to obtain. We collected a list of 75 network advertisers offering opt-out cookies that can be set using the Beef TACO (Targeted Advertising Cookie Opt-Out) Firefox extension. As detailed in
Appendix F, We found that only 11 of them delivered CPs
with their opt-out cookie. However, we found errors in only
one of these CPs.
Some of the most-visited domains host many websites in
their domain, and thus have many invalid CPs. Table 5 lists
the 6 domains responsible for at least 100 invalid CPs each.
These 6 domains are accountable for nearly 40% of CPs with
errors. These include two of the top 100 most-visited web
domains: tripod.com and msn.com.

7.

CONCLUSIONS

In this paper we present data on errors commonly found
in P3P compact policies that are detectable through automated analysis. We evaluated CPs collected from 33,139
websites on 19,820 domains and found invalid tokens, missing tokens, or conflicting tokens at 34% of these sites. We
found CP errors on a wide range of sites, including some of
the most popular websites on the Internet and TRUSTe seal
holders. We also reviewed the opt-out cookies of 75 network
advertisers, and found errors in one of the 11 CPs collected.
We were surprised by the large number of errors we were able
8

Host domain

Invalid CPs

vious work suggests that errors in full P3P policies are also
common [14] [3]. It appears that companies do not currently have sufficient incentives to provide accurate machinereadable privacy policies. Unless regulators use their authority to take action against companies that provide erroneous
machine-readable policies, users will be unable to rely on
these policies.

Percent of
total invalid CPs

tripod.com

2,575

23.0%

addresses.com

1,054

9.4%

msn.com

358

3.2%

cjb.net

247

2.2%

livedoor.biz

116

1.0%

112

1.0%

8.

39.92%

[1] S. Byers, L. F. Cranor, and D. Kormann. Automated
analysis of P3P-enabled web sites. ICEC ’03:
Proceedings of the 5th international conference on
Electronic commerce, pages 326–338, 2003.
[2] L. F. Cranor. Web Privacy with P3P. O’Reilly, 2002.
[3] L. F. Cranor, S. Egelman, S. Sheng, A. M. McDonald,
and A. Chowdhury. P3P deployment on websites.
Electronic Commerce Research and Applications,
7(3):274–293, 2008.
[4] L. F. Cranor, P. Guduru, and M. Arjula. User
interfaces for privacy agents. ACM Transactions on
Computer-Human Interaction, pages 135–178, June
2006.
[5] S. Hacker. P3P in IE: Frustrating failure, June 2002.
http://www.oreillynet.com/mac/blog/2002/06/
p3p_in_ie6_frustrating_failure.html.
[6] J. A. Harvey and K. M. Sanzaro. P3P and IE 6: good
privacy medicine or mere placebo? The Computer and
Internet Lawyer, 2002.
[7] http://www.w3.org/2002/12/18-p3p-workshop
report.html. Summary report - W3C workshop on the
future of P3P. Technical report, W3C, November
2002.
[8] P. G. Kelley, J. Bresee, L. F. Cranor, and R. W.
Reeder. A “nutrition label” for privacy. SOUPS ’09:
Proceedings of the 5th Symposium on Usable Privacy
and Security, 2009.
[9] Microsoft. Privacy in Internet Explorer 6. Visited
April 26, 2010.
http://msdn.microsoft.com/en-us/library/ms537343
[10] Microsoft Support. Session variables are lost if you use
FRAMESET in Internet Explorer 6, April 2006.
http://support.microsoft.com/kb/323752.
[11] G. Patel. How to set third-party cookies with iframe,
December 2008.
http://viralpatel.net/blogs/2008/12/how-to-set-thirdparty-cookies-with-iframe.html.
[12] K. Patil. Session lost in iframe, April 2010.
http://kiranpatils.wordpress.com/2010/04/13/sessionlost-in-iframe/.
[13] Piskvor. Cookie blocked/not saved in IFRAME in IE,
February 2009.
http://stackoverflow.com/questions/389456/cookieblocked-not-saved-in-iframe-in-internet-explorer.
[14] I. Reay, S. Dick, and J. Miller. A large-scale empirical
study of P3P privacy policies: Stated actions vs. legal
obligations. ACM Transactions on the Web, 2009.
[15] J. R. Reidenberg and L. F. Cranor. Can user agents
accurately represent privacy policies? The 38th
Research Conference on Communication, Information
and Internet Policy, 2002.
[16] TRUSTe. Directory of TRUSTe certified web sites.

ning.com
Total invalid:

4,462

Table 5: Domains accountable for most of the CP
problems.

to detect in CPs through automated analysis alone. We expect that even more errors exist, but discovering them would
require manual comparison with sites’ human-readable privacy policies or first-hand knowledge of sites’ actual privacy
practices. The large number of CP errors is troubling and
suggests that CPs cannot be relied on for accurate information about website privacy policies with respect to cookies.
We conducted a number of analyses to try to understand
why such a large fraction of CPs contain errors. Our results
suggest that while some errors are likely introduced through
mistakes (e.g. typos or misunderstanding the P3P specification), most appear to result from web administrators writing
CPs for the purpose of avoiding IE cookie filtering without
considering the accuracy of their CPs. In addition, we found
large numbers of websites sharing the same erroneous CPs,
including groups of websites hosted on the same domain.
P3P is designed to provide website privacy policies in a
computer-readable format that enables automated analysis
and decision making. CPs provide a simple way for websites
to offer a summary of their privacy practices with respect to
cookies in a format that is easily processed by web browsers.
The IE web browser uses CPs to make cookie blocking decisions. Thus, CP errors are likely to cause IE to allow cookies
that should be blocked under a user’s privacy settings to go
unblocked, and users who rely on IE’s cookie settings may
be misled. This problem is exacerbated by the fact that the
IE cookie-filtering implementation does not check for CP errors. Thus even the invalid- and missing-token errors, which
are a clear violation of the P3P specification, go undetected
by IE. Indeed, some websites appear to exploit this IE implementation loophole and publish CPs containing only bogus tokens or omitting tokens in the categories that would
cause IE to filter their cookies. A number of online articles
also suggest CPs that websites can use to avoid having their
cookies blocked, and we found large numbers of sites that
copied these suggested CPs verbatim.
CP errors would likely be reduced substantially if IE checked
for these errors, and if the articles that informed web administrators about avoiding cookie blocking explained that CPs
need to follow the P3P specification and accurately represent privacy practices with respect to cookies. In addition,
if the administrators of domains that host large numbers of
websites corrected the CPs for their domains, the number of
errors would be significantly reduced.
The CP error data we report suggests that many websites are not taking P3P seriously and are behaving in ways
that undermine the purpose of the P3P specification. Pre9

REFERENCES

Visited May 3, 2010.
http://www.truste.com/trusted sites/index.html.
[17] W3C. The platform for privacy preferences 1.0, April
2002. http://www.w3.org/TR/REC-P3P-20020416.
[18] W3C. The platform for privacy preferences 1.1.
http://www.w3.org/TR/P3P11/, November 2006.
[19] A. Young. IE blocking iframe cookies, September 2008.
http://adamyoung.net/IE-Blocking-iFrame-Cookies.

10

APPENDIX
A.

DESCRIPTION OF P3P COMPACT POLICY TOKENS

Element

Token

Full P3P Vocabulary

Plain Language Translation of P3P Policy Element [18]

Access

NOI
ALL
CAO

<nonident/>
<all/>
<contact-and-other/>

IDC

<ident-contact/>

OTI

<other-ident/>

NON

<none/>

We do not keep any information identified with you
We give you access to all of our information identified with you
We give you access to your contact information and some of our
other information identified with you
We give you access to only your contact information in our
records
We allow you to access some of our information identified with
you, but not your contact information
We do not give you access to our information about you

Disputes

DSP

There are some disputes

There are ways to resolve privacy-related disputes with us

Remedies

COR

<correct/>

MON

<money/>

LAW

<law/>

We will correct any errors we make related to the commitments
in our privacy policy
We will compensate individuals if it is determined that we have
violated our privacy policy
Our privacy policy references a law that may determine remedies
for breaches of our policy

NonIdentifiable

NID

<NONIDENTIFIABLE/>

We do not keep any information that could be used to identify
you personally

Purpose

CUR
ADM[attr]
DEV[attr]

<current/>
<admin/>
<develop/>

TAI[attr]
PSA[attr]

<tailoring/>
<pseudo-analysis/>

PSD[attr]

<pseudo-decision/>

IVA[attr]
IVD[attr]

<individual-analysis/>
<individual-decision/>

CON[attr]

<contact/>

HIS[attr]

<historical/>

TEL[attr]
OTP[attr]

<telemarketing/>
<other-purpose/>

To provide the service you requested
To perform web site and system administration
For research and development, but without connecting any information to you
To customize the site for your current visit only
To do research and analysis in which your information may be
linked to an ID code but not to your personal identity
To make decisions that directly affect you without identifying
you, for example to display content or ads based on links you
clicked on previously
To do research and analysis that uses information about you
To make decisions that directly affect you using information
about you, for example to recommend products or services based
on your previous purchases
To contact you through means other than telephone (for example, email or postal mail) to market services or products
To aid in historical preservation as governed by a law or policy
described in this privacy policy
To contact you by telephone to market services or products
For other uses described in the site’s human readable policy

OUR

<ours/>

DEL[attr]

<delivery/>

SAM[attr]
UNR[attr]
PUB[attr]

<same/>
<unrelated/>
<public/>

OTR[attr]

<other-recipient/>

Recipient

Companies that help us fulfill your requests (for example, shipping a product to you), but these companies must not use your
information for any other purpose
Delivery companies that help us fulfill your requests and who
may also use your information in other ways
Companies that have privacy policies similar to ours
Other companies whose privacy policies are unknown to us
People who may access your information from a public area,
such as a bulletin board, chat room, or directory
Companies that are accountable to us, though their privacy policies may be different from ours
Continued on next page . . .

11

Continued from previous page

Element

Token

Full P3P Vocabulary

Description

Retention

NOR

<no-retention/>

STP

<stated-purpose/>

LEG

<legal-requirement/>

BUS

<business-practices/>

IND

<indefinitely/>

We do not keep your information beyond your current online
session
We keep your information only long enough to perform the activity for which we collected it
We keep your information only as long as we need to for legal
purposes
Our full privacy policy explains how long we keep your information
We may keep your information indefinitely

PHY

<physical/>

ONL
UNI

<online/>
<uniqueid/>

PUR
FIN

<purchase/>
<financial/>

COM

<computer/>

NAV

<navigation/>

INT

<interactive/>

DEM

<demographic/>

CNT

<content/>

STA
POL

<state/>
<political/>

HEA

<health/>

PRE
LOC

<preference/>
<location/>

GOV
OTC

<government/>
<other-category/>

Name, address, phone number, or other physical contact information
Email address or other online contact information
Website login IDs and other identifiers (excluding government
IDs and financial account numbers)
Information about your purchases, including payment methods
Financial information such as accounts, balances, and transaction history
Information about the computer you are using, such as its hardware, software, or Internet address
Which pages you visited on this web site and how long you
stayed at each page
Activities you engaged in at this web site, such as your searches
and transactions
Information about social and economic categories that might
apply to you, such as your gender, age, income, or where you
are from
Messages you send to us or post on this site, such as email,
bulletin board postings, or chat room conversations
Cookies and mechanisms that perform similar functions
Which groups you might be a member of such as religious organizations, trade unions, and political parties
Health information such as information about your medical condition or your interest in health-related topics, services, or products
Information about your tastes or interests
Information about an exact geographic location, such as data
transmitted by your GPS-enabled device
Government-issued identifiers such as social security numbers
Other types of data described in the site’s human readable policy

TST

<test/>

The CP is under test

Categories

Test

Attributes [attr]: a = always, i = opt-in, o = opt-out

12

B.

FREQUENCY OF TOKENS

Element

Token

always

opt-in

opt-out

Access

NOI
ALL
CAO
IDC
OTI
NON

28%
6%
44%
12%
<1%
5%

NA
NA
NA
NA
NA
NA

NA
NA
NA
NA
NA
NA

Disputes

DSP

61%

NA

NA

Remedies

COR
MON
LAW

54%
<1%
3.4%

NA
NA
NA

NA
NA
NA

NonNID
identifiable

6%

NA

NA

Purpose

CUR/CURa
ADM
DEV
TAI
PSA
PSD
IVA
IVD
CON
HIS
TEL
OTP

58%
73%
70%
37%
52%
32%
11%
2%
1%
3%
<1%
1%

NA
1%
1%
2%
16%
1%
27%
23%
30%
<1%
1%
22%

NA
1%
1%
<1%
2%
1%
2%
3%
12%
<1%
22%
<1%

Recipient

OUR
DEL
SAM
UNR
PUB
OTR

96%
3%
1%
<1%
<1%
2%

NA
22%
22%
21%
21%
23%

NA
<1%
2%
<1%
<1%
15%

Retention

NOR
STP
LEG
BUS
IND

7%
20%
1%
13%
67%

NA
NA
NA
NA
NA

NA
NA
NA
NA
NA

Categories

PHY
ONL
UNI
PUR
FIN
COM
NAV
INT
DEM
CNT
STA
POL
HEA
PRE
LOC
GOV
OTC

39%
40%
56%
31%
23%
61%
61%
39%
49%
25%
45%
21%
21%
31%
15%
21%
3%

NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

Test

TST

<1%

NA

NA

This table shows the percentage of compact policies in our
data set that contain each CP token.

13

C.

EVALUATION OF FULL P3P AND HUMAN-READABLE POLICIES FOR WEB SITES
USING THE CP SUGGESTED BY THE O’REILLY BLOG

URL

Valid
Full
P3P
Policy?

Location of
Human-readable
Policy

Comments

alleghenyinstitute.org

NO

Not found

No policies found to compare with CP

bordellfuehrer.de

NO

Not found

No policies found to compare with CP

caidep.com

NO

caidep.com

Policies do not match - privacy policy does not mention any information associated with the N OI, ADM ,
DEV , P SAi, COM , OT Ro, ST P , IN D or DEM tokens
included in the CP; privacy policy mentions the use of
cookies to store preferences and to perform customization
but CP does not include P RE or T AI tokens

cakephp.org

NO

cakephp.org/pages/privacy

Policies do not match - privacy policy does not mention any information associated with the N OI, ADM ,
DEV , P SAi, ST P , IN D or DEM tokens included in its
CP; privacy policy mentions the use of cookies to store
preferences but CP does not include P RE token

campbell.house.gov

NO

Not found

No policies found to compare with CP

condusef.gob.mx

NO

Not found

No policies found to compare with CP

creditolo.de

NO

Not found

No policies found to compare with CP

dme.kerala.gov.in

NO

Not found

No policies found to compare with CP

economics.harvard.edu NO

Not found

No policies found to compare with CP

equestrian.com.my

NO

equestrian.com.my/
privacy-policy

Human-readable policy does not mention cookies

gilldivers.com

NO

Not found

No policies found to compare with CP

gss.ucsb.edu

NO

Not found

No policies found to compare with CP

honor.unc.edu

NO

Not found

No policies found to compare with CP

itech-ny.com

NO

itech-ny.com/
privacy-policy.html

Policies do not match - privacy policy does not mention any information associated with N OI, ADM , DEV ,
ST P , IN D or DEM tokens included in its CP; privacy
policy mentions the use of cookies to store preferences but
CP does not include the P RE token

joomla.org

NO

joomla.org/
privacy-policy.html

Human-readable policy does not mention cookies

komodorock.com

NO

www.komodorock.com/
privacy-policy/

Policies do not match - privacy policy does not mention any information associated with N OI, ADM , DEV ,
ST P , IN D or DEM tokens included in CP; privacy policy mentions the use of cookies to store preferences and
customize advertising but CP does not include the P RE
or T AI tokens

laser.org

NO

Not found

No policies found to compare with CP

majorleague.com.au

NO

Not found

No policies found to compare with CP

megasearch.net

NO

megasearch.net/
PrivacyPolicy.html

Policies do not match - the use of cookies is not well
detailed in the human-readable policy.
Continued on next page . . .

14

Continued from previous page

URL

Valid
Full
P3P
Policy?

Location of
Human-readable
Policy

Comments

navicat.com

NO

navicat.com/en/privacy.html Human-readable policy does not mention cookies

ocean.tamu.edu

NO

geosciences.tamu.edu/
about-us/
geonet-information-hub/
web-site-policies/
677-site-privacy-andsecurity-policy

Policies do not match - privacy policy does not mention any information associated with N OI, DEV , P SAi,
OT Ro, ST P , IN D or DEM tokens included in CP

orange-pocket.com

NO

Not found

No policies found to compare with CP

parktrust.org

NO

Not found

No policies found to compare with CP

rcn.com

NO

rcn.com/dc-metro/
privacy-policy

Policies do not match - the use of cookies is not well detailed in the human-readable policy - privacy policy does
not mention any information associated with N OI, ADM ,
DEV , P SAi, ST P , IN D or DEM tokens included in CP;
privacy policy mentions that cookies are used to provide
seamless visit and expedite customer login but CP does
not include CU R token

relevantmagazine.com

NO

relevantmagazine.com/
privacy-policy

Policies do not match - the use of cookies is not well
detailed in the human-readable policy

themacstore.com

NO

themacstore.com/privacy/

Human-readable policy does not mention cookies

theories.com

NO

theories.com/index.php/
Privacy-Policy.html

Policies do not match - the use of cookies is not well
detailed in the human-readable policy

topcities.com

NO

Not found

No policies found to compare with CP

womensmedia.com

NO

womensmedia.com/new/
privacy-policy.shtml

Human-readable policy does not mention cookies

wsashow.com

NO

wsashow.com/homepage/
privacy policy

Policies do not match - privacy policy does not mention any information associated with N OI, DEV , ST P
or IN D tokens included in its CP; privacy policy mentions that cookies are used for log-in, enable personalization, analytics, shopping cart, personalized service, and
targeted advertisement, but CP does not include CU R,
IV D, IV A, T AI or IN T tokens

15

D.

EVALUATION OF FULL P3P AND HUMAN-READABLE POLICIES FOR WEBSITES WITH
CP ERRORS IN TOP 50 MOST-VISITED LIST

URL

Valid
Full
P3P
Policy?

Location of
Human-readable
Policy

Comments

facebook.com

NO

facebook.com/policy.php

Policies do not match - CP contains only two tokens
(DSP and LAW ) but privacy policy mentions that cookies are used for several purposes, including the provision of
services, advertising, easy log-in, etc., and that cookies are
stored for an extended period

msn.com

NO

privacy.microsoft.com/
en-us/fullnotice.mspx

Slight differences between CP and privacy policy privacy policy mentions that cookies may be used to collect
demographic information but CP does not include DEM
token

safety.live.com

NO

privacy.microsoft.com/
en-us/fullnotice.mspx

Slight differences between CP and privacy policy privacy policy mentions that cookies may be used to collect
demographic information but CP does not include DEM
token

amazon.com

NO

amazon.com/gp/help/
customer/display.html
/ref=footer privacy/
191-3583711-6331321?
ie=UTF8&nodeId=468496

Invalid CP, unable to compare

microsoft.com

YES

privacy.microsoft.com/
en-us/fullnotice.mspx

Slight differences between CP, full P3P policy, and
privacy policy - privacy policy mentions that cookies may
be used to collect demographic information but CP and full
P3P policy do not include DEM token

reference.aol.com

NO

about.aol.com/aolnetwork/
aol pp

Policies do not match - privacy policy mentions cookies
are used to remember preferences, measure ad effectiveness,
customize site, store demographic information, share info
with ad networks and service providers, but CP does not
include any P RE, ON L, T AI, DEM or SAM tokens.

atlas.mapquest.com

NO

about.aol.com/aolnetwork/
aol pp

Policies do not match - privacy policy mentions cookies
are used to remember preferences, measure ad effectiveness,
customize site, store demographic information, share info
with ad networks and service providers, but CP does not
include any P RE, ON L, T AI, DEM or SAM tokens.

godaddy.com

NO

godaddy.com/
Agreements/ShowDoc.aspx?
pageid=PRIVACY&ci=
20803&app hdr=0

Policies do not match - privacy policy mentions the collection of name, address, credit card numbers, government
IDs, and collected information might be used to contact the
user and to present co-branded offers on opt-in basis, but
CP does not contain P HY , DEM , GOV , CON or SAM o

imdb.com

NO

imdb.com/privacy

Invalid CP, unable to compare

windows.com

NO

privacy.microsoft.com/
en-us/fullnotice.mspx

Slight differences between CP and privacy policy privacy policy mentions that cookies may be used to collect
demographic information but CP does not include DEM

hulu.com

NO

hulu.com/privacy

Policies do not match - privacy policy mentions targeted
advertising based on user’s activity but CP does not include
IV A; policy states: “We may use cookies and similar technologies to relate your use of the Hulu Services to personally
identifiable information,” yet CP includes the N ID token,
claiming that they do not collect PII

16

E.

TOP VISITED DOMAINS USING CPS
Domain

Valid Full
P3P Policy?

Errors found
in CP

about.com
amazon.com
angelfire.com
aol.com
apple.com
att.com
bing.com
bizrate.com
blogspot.com
careerbuilder.com
causes.com
cnet.com
cnn.com
comcast.net
dailymotion.com
examiner.com
facebook.com
flickr.com
go.com
godaddy.com
google.com
hulu.com
ign.com
imdb.com
latimes.com
linkedin.com
live.com
mapquest.com
match.com
metacafe.com
microsoft.com
monster.com
msn.com
mybloglog.com
nytimes.com
people.com
simplyhired.com
target.com
thefind.com
tripod.com
tumblr.com
twitter.com
washingtonpost.com
weatherbug.com
wikipedia.org
windows.com
yahoo.com
yellowpages.com

YES
NO
NO
NO
NO
YES
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
YES
NO

None
Invalid tokens; Missing tokens
Invalid tokens; IVA and CON conflicting tokens
Missing tokens
None
None
None
Invalid tokens
None
IVA conflicting token
IVD, IVA and CON conflicting tokens
None
None
None
None
NID conflicting token
Missing tokens
None
Invalid tokens; Missing tokens
IVD and IVA conflicting tokens
None
Invalid tokens
None
Invalid tokens; Missing tokens
None
None
Missing tokens
Missing tokens
None
None
CON conflicting token
None
Invalid tokens
None
Missing tokens
None
None
None
IVD and IVA conflicting token
Invalid tokens; IVA and CON conflicting tokens
None
None
TEL, IVD, IVA, and CON conflicting tokens
None
None
Invalid tokens
None
None

Total

7/48

21/48

17

F.

NETWORK ADVERTISING DOMAINS USING CPS
Domain

Errors found
in CP

247realmedia.com
adsfac.sg
atdmt.com
casalemedia.com
imiclk.com
intellitxt.com
navegg.com
realmedia.com
vizu.com
weborama.fr
zedo.com

YES
YES
YES
NO
YES
NO
NO
YES
YES
NO
NO

None
None
None
None
None
None
Invalid tokens
None
None
None
None

Total

G.

Valid Full
P3P Policy?

6/11

1/11

DOMAINS HOLDING TRUSTE SEALS USING CPS
Domain

Valid Full
P3P Policy?

Errors found
in CP

10kscholarship.com
1800mobiles.com
192.com
1choice4yourstore.com
247realmedia.com
2fixyourtrafficticket.com
3dcart.com
abc.com
abcnews.com
activeinternational.ca
activeinternational.com
adt.com
agilent.com
alladvertisingagencies.com
aloharents.com
alvenda.com
amiastri.com
angelfire.com
aol.com
apothica.com
appexchange.com
apple.com
asksanta.ca
att.com
att.net
attinteractive.com
automationcontrols.com
autonation.com
avaline.com
aviationarthangar.com
bellsouth.com
bic-gsa.com
bicgsa.com

NO
NO
YES
NO
YES
YES
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
NO
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO

None
None
None
None
None
None
None
None
None
IVD, IVA, and CON conflicting tokens
IVD, IVA, and CON conflicting tokens
None
Missing tokens
None
None
None
Invalid tokens
Invalid tokens; IVA and CON conflicting tokens
Missing tokens
Missing tokens; TEL and CON conflicting tokens
Missing tokens
None
Missing tokens
None
Invalid tokens
None
None
None
None
None
None
None
None
Continued on next page . . .

18

Continued from previous page

Domain

Valid Full
P3P Policy

Errors found
in CP

bicwarehouse.com
bidezone.com
billhighway.com
billiardsaddiction.com
billshrink.com
bing.com
bizrate.com
bizrate.de
bluerazor.com
bodymedia.com
boostflow.com
boston.com
burstnet.com
buyingadvice.com
buysafe.com
buysafeshopping.com
caliberlocal.com
calibex.com
candlewoodsuites.com
candywarehouse.com
caoh.org
careonecredit.com
carid.com
casalemedia.com
caspio.com
caspio.net
cataloglink.com
catchfirefunding.com
cellstores.com
ceu4u.com
chatthreads.com
cheaptickets.com
chefsresource.com
chegg.com
chipin.com
christmastreeforme.com
cjhomeandoffice.com
classmates.com
clcleather.net
clubbing.com
code7contest.com
comcast.net
conair-store.com
concreteexchange.com
controlscan.com
coremetrics.com
costumecity.com
couponbug.com
coupons.com
couponsinc.com
cpp.com
crafta.com
credit.com
criteo.com

NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
YES
NO
NO
NO
YES
YES
YES
NO
NO
NO
NO
NO
NO
NO

None
Missing tokens
Invalid tokens; Missing tokens
None
IVD, IVA, and CON conflicting tokens
None
Invalid tokens
Invalid tokens
IVD and IVA conflicting tokens
None
None
None
None
None
IVD, IVA, and CON conflicting tokens
IVD and IVA conflicting tokens
None
None
Missing tokens
None
None
None
None
None
Missing tokens
Missing tokens
NID conflicting token
None
None
None
Missing tokens
IVD and IVA conflicting tokens
None
None
None
None
None
IVD, IVA, and CON conflicting tokens
None
Missing tokens
None
None
None
None
None
None
None
None
None
None
Missing tokens
None
None
None
Continued on next page . . .

19

Continued from previous page

Domain

Valid Full
P3P Policy

crowneplaza.com
NO
cufflinksdepot.com
NO
datepad.com
NO
dealsonhotels.com
NO
debtgoal.com
NO
depositagift.com
NO
dexclusive.com
NO
digicert.com
YES
digilifestudios.com
NO
digitalimaginghq.com
NO
digitallanding.com
NO
digitalspyders.com
NO
directfix.com
NO
directtextbook.com
YES
dreamlandweddingshoppe.com NO
drugs.com
YES
duiattorney.com
NO
dynamiclogic.com
YES
e-miles.com
NO
e-rewards.com
NO
e-rewards.de
NO
e-rewards.fr
NO
e-rewards.nl
NO
ea.com
NO
earnmydegree.com
NO
ebates.com
NO
ebooks.com
YES
ecampustours.com
NO
echosign.com
NO
educadium.com
NO
educationconnection.com
YES
emeraldisland.com
NO
emergingmed.com
NO
enjoycpr.com
NO
epals.com
NO
eprooft.com
NO
espn.com
YES
ether.com
NO
eversave.com
NO
facebook.com
NO
familyfun.com
NO
fansnap.com
NO
federaldebtreduction.com
NO
firstagain.com
NO
flemingoutdoors.com
NO
forzieri.com
NO
freeshop.com
NO
genealogytoday.com
NO
getaroom.com
NO
getinsurancequotes.ca
NO
globesmart.com
NO
go.com
YES
godaddy.com
NO
gotomypc.com
NO

Errors found
in CP
Missing tokens
None
Missing tokens
Missing tokens
None
None
None
CON conflicting token
None
None
IVD, IVA, and CON conflicting tokens
None
None
None
None
None
None
None
None
None
None
None
None
Missing tokens
None
None
Missing tokens; CON conflicting token
None
IVD, IVA, and CON conflicting tokens
None
IVD, IVA, and CON conflicting tokens
None
None
None
None
None
None
None
None
Missing tokens
None
None
None
CON conflicting token
None
None
NID conflicting token
None
None
None
None
Invalid tokens; Missing tokens
IVD and IVA conflicting tokens
Missing tokens
Continued on next page . . .

20

Continued from previous page

Domain

Valid Full
P3P Policy

Errors found
in CP

gowearfit.com
greenfieldonline.com
greensherpa.com
greenwayuniversity.com
grovesite.com
healthscout.com
healthsquare.com
hiexpress.com
higherone.com
holiday-inn.com
homedecorhardware.com
homegain.com
hotbot.com
hotelindigo.com
houstontexans.com
htmlgear.com
hyperstreet.com
ibm.com
ichotelsgroup.com
ideascale.com
ifriends.net
ifriendsv2.net
ihg.com
ihgarmyhotels.com
importedblankets.com
inksell.com
inoutcash.com
insightexpress.com
intelius.com
intercontinental.com
intuit.com
itech-ny.com
itwixie.com
jackpotrewards.com
jaman.com
jameslimousines.com
jewelrywonder.com
jobtarget.com
justasktoday.com
kanetix.ca
kanetix.com
karmacar.com
keen.com
keysurvey.com
kinglinen.com
largestmall.com
legalmatch.com
letstalk.com
life360.com
lifequote.com
linkedin.com
listyourdebt.com
lithium.com
live.com

NO
NO
NO
NO
NO
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO

None
None
None
None
None
None
None
Missing tokens
None
Missing tokens
None
None
CON conflicting token
Missing tokens
None
Invalid tokens; IVA and CON conflicting tokens
None
None
Missing tokens
None
None
None
Missing tokens
Missing tokens
None
None
None
None
Missing tokens
Missing tokens
Missing tokens
None
None
Invalid tokens
Missing tokens
None
Missing tokens
Invalid tokens; Missing tokens
None
None
None
None
None
Missing tokens
None
Missing tokens
None
None
None
None
None
None
None
Missing tokens
Continued on next page . . .

21

Continued from previous page

Domain

Valid Full
P3P Policy

Errors found
in CP

livemeeting.com
loanio.com
logcap4jobs.com
lycos.com
maghound.com
mail2world.com
mailchimp.com
market2lead.com
mate1.com
maven.net
mba.com
mcmobileaccessories.com
medelita.com
medlink.com
medsurvey.com
mercedsystems.com
mesh.com
microsoft-hohm.com
microsoft.com
microsoftfinancing.com
microsofthohm.com
mitto.com
mndigital.com
moneybookers.com
monster.ch
monster.com
moversdeal.com
msn.at
msn.be
msn.com
msn.de
msn.dk
msn.es
msn.fi
msn.fr
msn.it
msn.nl
msn.no
msn.pt
msn.se
mybarstools.com
myfreepaysite.com
myhomepage.com
mynewplace.com
napster.com
napster.de
nationalgamecity.com
nextag.ca
nextag.com
nflflag.com
nupplegal.com
nytimes.com
oakcitygallery.com
officedrop.com

NO
NO
NO
NO
NO
NO
NO
YES
NO
YES
NO
NO
NO
YES
NO
NO
NO
NO
YES
NO
YES
NO
NO
YES
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO

CON conflicting token
None
None
CON conflicting token
None
Missing tokens; IVA and CON conflicting tokens
Invalid tokens; Missing tokens
None
None
None
None
None
None
None
None
None
Missing tokens
None
CON conflicting token
None
None
CON conflicting token
Invalid tokens
None
None
None
None
Missing tokens
Missing tokens
Invalid tokens
Missing tokens
Missing tokens
Missing tokens
Missing tokens
Missing tokens
Missing tokens
Missing tokens
Missing tokens
Missing tokens
Missing tokens
None
NID conflicting token
CON conflicting token
Missing tokens; CON conflicting token
None
None
None
None
None
None
None
Missing tokens
None
None
Continued on next page . . .

22

Continued from previous page

Domain

Valid Full
P3P Policy

Errors found
in CP

omniture.com
onebagoneearth.com
onesky.com
onetravel.com
onetravelindia.com
onewayfurniture.com
opinion-central.com
orbitz.com
orbitzforbusiness.net
paybycash.com
paycycle.com
payscale.com
pch.com
pcicomplianceguide.org
pensxpress.com
peopleclick.com
perfectmatch.com
periogen.com
permuto.com
photosynth.net
pictureyoursunique.com
pinnaclesys.com
platinumgalleria.com
popularmedia.com
posonlinestore.com
pospaper.com
precharge.com
predictiveresponse.com
press8.com
priorityclub.com
priortax.com
prixmoinscher.fr
prodebtsupport.com
prosperitypublications.net
qualityhealth.com
quickenbillpay.com
quikcondoms.com
racingusa.com
rapidrefund.net
rapidrepair.com
rapidtax.com
rednel.com
remington-store.com
rent.com
repequity.com
rewardtv.com
rixty.com
roblox.com
rockstargames.com
rockyou.com
rozee.pk
safecount.net
salesforcefoundation.org
sharefile.com

NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
NO
YES
NO
NO
NO
NO
YES
NO
NO
NO
YES
NO
NO
NO
YES
NO
NO
NO
YES
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO

None
None
None
None
None
None
None
IVD and IVA conflicting tokens
IVD and IVA conflicting tokens
IVD, IVA, and CON conflicting tokens
Missing tokens
IVD, IVA, and CON conflicting tokens
None
None
None
None
None
None
None
Invalid tokens
None
None
None
None
None
None
None
None
None
Missing tokens
Invalid tokens; Missing tokens
None
None
None
None
None
None
None
Invalid tokens; Missing tokens
None
Invalid tokens; Missing tokens
None
None
None
None
IVD and IVA conflicting tokens
None
None
Missing tokens
None
None
None
Missing tokens
Missing tokens
Continued on next page . . .

23

Continued from previous page

Domain

Valid Full
P3P Policy

Errors found
in CP

shermanstravel.com
shop.com
shopbrita.com
shopcompanion.com
shopdeck.com
shopiogear.com
shopkitchenaid.com
shopzilla.com
shopzilla.de
shopzilla.fr
shustir.com
simplifi.net
simplybabyfurniture.com
simplykidsfurniture.com
sixcontinentsclub.com
skincarerx.com
skintreatment.com
smartsourceonline.com
snaglo.com
snapfish.com
soccernet.com
spardeingeld.de
spendgrowgive.com
spiceworks.com
spoke.com
spokesoftware.com
sportingnews.com
spytown.com
starfieldtech.com
starwars.com
staybridge.com
strands.com
suresource.com
surveillance-video.com
sweatmonkey.org
talentfilter.biz
taxact.com
taxactonline.com
taxcut.com
taxpack.com
techbargains.com
techcctv.com
theblueriverbabyshoppe.com
theopenskyproject.com
thesims2.com
thumbplay.com
toluna.com
topdjgear.com
toponeshop.com
torbalscales.com
treadmilldoctor.com
tripit.com
tripod.com
tycoonu.com

NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
NO
YES
NO
YES
NO
YES
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO

None
NID conflicting token
None
NID conflicting token
None
None
None
Invalid tokens
Invalid tokens
Invalid tokens
None
None
None
None
Missing tokens
Missing tokens; TEL and CON conflicting tokens
None
None
None
None
None
Invalid tokens
None
None
IVD, IVA, and CON conflicting tokens
IVD, IVA, and CON conflicting tokens
None
None
IVD and IVA conflicting tokens
Missing tokens
Missing tokens
None
None
None
None
IVD, IVA, and CON conflicting tokens
IVA and CON conflicting tokens
IVA and CON conflicting tokens
None
Invalid tokens; Missing tokens
IVD, IVA, and CON conflicting tokens
None
None
None
Missing tokens
IVD, IVA, and CON conflicting tokens
None
None
None
None
None
Missing tokens
Invalid tokens; IVA and CON conflicting tokens
Missing tokens
Continued on next page . . .

24

Continued from previous page

Domain

Valid Full
P3P Policy

Errors found
in CP

ultimatepay.com
unbeatablesale.com
unique-egifts.com
us-appliance.com
uscretailproducts.com
verisign.com
verizon.net
vermontgear.com
viewpoint.com
vitadigest.com
vitamaker.com
voice123.com
w3i.com
wallpapers.com
waterpik-store.com
weatherbug.com
webtv.net
westfloridacomponents.com
whitakertaylor.com
whitesmoke.com
whowhere.com
wildwestdomains.com
windowsmedia.net
wine.com
winferno.com
wirefly.com
wirelessground.com
wisemanfinance.com
wondertime.com
wooddashexperts.com
xbox.com
yahoo.com
yellowpages.com
zocdoc.com

NO
YES
NO
NO
NO
NO
NO
NO
YES
NO
NO
YES
NO
NO
NO
YES
NO
NO
YES
YES
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO

IVD, IVA, and CON conflicting tokens
None
None
None
None
None
IVD, IVA, and CON conflicting tokens
None
None
None
None
None
Invalid tokens; Missing tokens
Invalid tokens; Missing tokens
None
None
Missing tokens
None
None
Missing tokens
CON conflicting token
IVD and IVA conflicting tokens
None
TEL and CON conflicting tokens
None
None
None
None
None
None
Invalid tokens
None
None
Missing tokens

Total

63/391

134/391

25