text
stringlengths 195
512
|
|---|
ple, it only provides a high-level methodology instead
of the exact solution to user’s question.
3: It means the answer is helpful but not written by an AI Assistant. It addresses all
the basic asks from the user. It is complete and self contained with the drawback that
the response is not written from an AI assistant’s perspective, but from other people’s
perspective. The content looks like an excerpt from a blog post, web page, or web search
results.
For example, it contains personal experience or opinion |
, mentions comments
section, or share on social media, etc.
4: It means the answer is written from an AI assistant’s perspective with a clear focus of
addressing the instruction. It provide a complete, clear, and comprehensive response to
user’s question or instruction without missing or irrelevant information. It is well organized,
self-contained, and written in a helpful tone. It has minor room for improvement, e.g. more
concise and focused.
5: It means it is a perfect answer from an AI Assistant. It has |
a clear focus on being a
helpful AI Assistant, where the response looks like intentionally written to address the
user’s question or instruction without any irrelevant sentences. The answer provides high
quality content, demonstrating expert knowledge in the area, is very well written, logical,
easy-to-follow, engaging and insightful.
User: <INSTRUCTION_HERE>
<response><RESPONSE_HERE></response>
Please first briefly describe your reasoning (in less than 100 words),
and then
write “Score: <rating>” in the la |
st line. Answer in the style of an AI Assistant, with
knowledge from web search if needed. To derive the final score based on the criteria, let’s
think step-by-step.
Figure 5: LLM-as-a-Judge prompt taken from Li et al. [2023a].
15
|
Data Sanitization Techniques
A Net 2000 Ltd. White Paper
Abstract
Data Sanitization is the process of making sensitive information in non-production databases
safe for wider visibility. This White Paper is an overview of various techniques which can be
used to sanitize sensitive production data in test and development databases.
An initial discussion of the primary motivations for data sanitization is given. The remainder of
the paper is devoted |
to a generic survey of the various masking techniques and their individual
benefits and drawbacks.
Some keywords which may assist you in finding this document online are:
Data Sanitization, Data Sanitisation, Data Masking, Data Obfuscation, Data Security, Data
Cleansing, Data Hiding, Data Protection Act 1998, Hide Data, Disguise Data, Sanitize Data,
Sanitise Data, Gramm-Leach-Bliley Act (GLBA), Data Privacy, Directive 95/46/EC of the
European Parliament
Author:
|
Dale Edgar
Net 2000 Ltd.
[email protected]
http://www.Net2000Ltd.com
Data Sanitization Techniques
A Net 2000 Ltd. White Paper
Copyright © Net 2000 Ltd. 2003-2004
http://www.Net2000Ltd.com
Table of Contents
Why Sanitize Information in Test and Development Databases? .............................................1
Protecting Valuable Information.................................................................... |
........................1
Legal Obligations ...................................................................................................................1
Data Sanitization Techniques.....................................................................................................2
Technique: NULL’ing Out .....................................................................................................2
Technique: Masking Data.................................................................. |
....................................2
Technique: Substitution .........................................................................................................3
Technique: Shuffling Records ................................................................................................4
Technique: Number Variance ................................................................................................4
Technique: Gibberish Generation ....................................................... |
..................................4
Technique: Encryption/Decryption........................................................................................5
Summary....................................................................................................................................6
Data Sanitization Techniques
A Net 2000 Ltd. White Paper
Copyright © Net 2000 Ltd. 2003-2004
http://www.Net2000Ltd.com
Data Sanitization Technique |
s
Data Sanitization is the process of disguising sensitive information in test and
development databases by overwriting it with realistic looking but false data of a
similar type.
Why Sanitize Information in Test and Development Databases?
The data in testing environments should be sanitized in order to protect valuable
business information and also because there is, in most countries, a legal obligation to
do so.
Protecting Valuable Information
Fundamentally there are two ty |
pes of security. The first type is concerned with the
integrity of the data. In this case the modification of the records is strictly controlled.
For example, you may not wish an account to be credited or debited without specific
controls and auditing. This type of security is not a major concern in test and
development databases. The data can be modified at will without any business impact.
The second type of security is the protection of the information content from
inappropriate visibility. Names |
, addresses, phone numbers and credit card details are
good examples of this type of data. Unlike the protection from updates, this type of
security requires that access to the information content is controlled in every
environment.
Legal Obligations
The legal requirements for Data Sanitization vary from country to country and most
countries now have regulations of some form. Here are some examples:
United States
The Gramm-Leach-Bliley Act requires institutions to protect the
confidential |
ity and integrity of personal consumer information. The Right to
Financial Privacy Act of 1978 creates statutory Fourth Amendment protection
for financial records and there are a host of individual state laws
The European Union
Directive 95/46/EC of the European Parliament which provides strict
guidelines regarding individual rights to data privacy, and the responsibilities
of data holders to guard against misuse.
Data Sanitization Techniques
A Ne |
t 2000 Ltd. White Paper
Copyright © Net 2000 Ltd. 2003-2004
http://www.Net2000Ltd.com
The United Kingdom
The United Kingdom Data Protection Act of 1998 extends the European
Parliament directive and places further statutory obligations on the holders of
personal, private or sensitive data.
As with most things legal, the details are open to argument. In reality though, if data
for which your organization is responsible gets loose and appropriate steps were not
taken to prevent that release, |
then your organizations lawyers could well find
themselves in court trying to put their best spin on the matter. However large the legal
liabilities are, they could seem trivial in comparison to the losses associated with the
catastrophic loss of business confidence caused by a large scale privacy breach.
Any organization that outsources test and development operations needs to be very
conscious of the specific laws regulating the transmission of information across
national borders.
Data Sa |
nitization Techniques
Test and development teams need to work with databases which are structurally
correct functional copies of the live environments. However, they do not necessarily
need to be able to view security sensitive information. For test and development
purposes, as long as the data looks real, the actual record content is usually irrelevant.
There are a variety of Data Sanitization techniques available – the pro’s and con’s of
some of the most useful are discussed below.
Technique: |
NULL’ing Out
Simply deleting a column of data by replacing it with NULL values is an effective
way of ensuring that it is not inappropriately visible in test environments.
Unfortunately it is also one of the least desirable options from a test database
standpoint. Usually the test teams need to work on the data or at least a realistic
approximation of it. For example, it is very hard to write and test customer account
maintenance forms if the customer name, address and contact details are all NULL |
values.
Verdict: The NULL’ing Out technique is useful in certain specific circumstances but
rarely useful as the entire Data Sanitization strategy.
Technique: Masking Data
Masking data means replacing certain fields with a Mask character (such as an X).
This effectively disguises the data content while preserving the same formatting on
front end screens and reports. For example, a column of credit card numbers might
look like:
Data Sanitization Techniques |
A Net 2000 Ltd. White Paper
Copyright © Net 2000 Ltd. 2003-2004
http://www.Net2000Ltd.com
4346 6454 0020 5379
4493 9238 7315 5787
4297 8296 7496 8724
and after the masking operation the information would appear as:
4346 XXXX XXXX 5379
4493 XXXX XXXX 5787
4297 XXXX XXXX 8724
The masking characters effectively remove much of the sensitive content from the
record while still preserving the look and feel. Take care to ensure that enough of the
data is masked to preserv |
e security. It would not be hard to regenerate the original
credit card number from a masking operation such as: 4297 8296 7496 87XX
since the numbers are generated with a specific and well known checksum algorithm.
Also care must be taken not to mask out potentially required information. A masking
operation such as XXXX XXXX XXXX 5379 would strip the card issuer details
from the credit card number. This may, or may not, be desirable.
Verdict: If the data is in a specific, invariable format, then Ma |
sking is a powerful and
fast Data Sanitization option. If numerous special cases must be dealt with then
masking can be slow, extremely complex to administer and can potentially leave some
data items inappropriately masked.
Technique: Substitution
This technique consists of randomly replacing the contents of a column of data with
information that looks similar but is completely unrelated to the real details. For
example, the surnames in a customer database could be sanitized by replacing the re |
al
last names with surnames drawn from a largish random list.
Substitution is very effective in terms of preserving the look and feel of the existing
data. The downside is that a largish store of substitutable information must be
maintained for each column to be substituted. For example, to sanitize surnames by
substitution, a list of random last names must be available. Then to sanitize telephone
numbers, a list of phone numbers must be available. Frequently, the ability to
generate known invalid |
data (phone numbers that will never work) is a nice-to-have
feature.
Substitution data can sometimes be very hard to find in large quantities. For example,
if a million random street addresses are required, then just obtaining the substitution
data can be a major exercise in itself.
Verdict: Substitution is quite powerful, reasonably fast and preserves the look and feel
of the data. Finding the required random data to substitute and developing the
procedures to accomplish the substitution can |
be a major effort.
Data Sanitization Techniques
A Net 2000 Ltd. White Paper
Copyright © Net 2000 Ltd. 2003-2004
http://www.Net2000Ltd.com
Technique: Shuffling Records
Shuffling is similar to substitution except that the substitution data is derived from the
column itself. Essentially the data in a column is randomly moved between rows until
there is no longer any reasonable correlation with the remaining information in the
row.
Th |
ere is a certain danger in the shuffling technique. It does not prevent people from
asking questions like “I wonder if so-and-so is on the supplier list?” In other words,
the original data is still present and sometimes meaningful questions can still be asked
of it. Another consideration is the algorithm used to shuffle the data. If the shuffling
method can be determined, then the data can be easily “unshuffled”. For example, if
the shuffle algorithm simply ran down the table swapping the column data i |
n between
every group of two rows it would not take much work from an interested party to
revert things to their unshuffled state.
Shuffling is rarely effective when used on small amounts of data. For example, if
there are only 5 rows in a table it probably will not be too difficult to figure out which
of the shuffled data really belongs to which row.
On the other hand, if a column of numeric data is shuffled, the sum and average of the
column still work out to the same amount. This can sometime |
s be useful.
Verdict: Shuffle rules are best used on large tables and leave the look and feel of the
data intact. They are fast and relatively simple to implement since no new data needs
to be found, but great care must be taken to use a sophisticated algorithm to
randomise the shuffling of the rows.
Technique: Number Variance
The Number Variance technique is useful on numeric data. Simply put, the algorithm
involves modifying each number value in a column by some random percentage of its
re |
al value. This technique has the nice advantage of providing a reasonable disguise
for the numeric data while still keeping the range and distribution of values in the
column within viable limits. For example, a column of sales data might have a
random variance of 10% placed on it. Some values would be higher, some lower but
all would be not too far from their original range.
Verdict: The number variance technique is occasionally useful and can prevent
attempts to correlate true records using known |
numeric data. This type of Data
Sanitization really does need to be used in conjunction with other options though.
Technique: Gibberish Generation
In general, when sanitizing data, one must take great care to remove all imbedded
references to the real data. For example, it is pointless to carefully remove real
customer names and addresses while still leaving intact in stored copies of
Data Sanitization Techniques
A Net 2000 Ltd. White Paper
|
Copyright © Net 2000 Ltd. 2003-2004
http://www.Net2000Ltd.com
correspondence in another table. This is especially true if the original record can be
determined via a simple join on a unique key.
Sanitizing “formless” non specific data such as letters, memos and notes is one of the
hardest techniques in Data Sanitization. Usually these types of fields are just
substituted with a random quantity of equivalently sized gibberish or random words.
If real looking data is required, either an elaborate |
substitution exercise must be
undertaken or a few carefully hand built examples must be judiciously substituted to
provide some representative samples.
Verdict: Occasionally it is useful to be able to substitute quantities of random text.
Gibberish Generation is useful when needed but is not a very widely applicable
technique.
Technique: Encryption/Decryption
This technique offers the option of leaving the data in place and visible to those with
the appropriate key while remaining effectivel |
y useless to anybody without the key.
This would seem to be a very good option – yet, as with all techniques, it has its
strengths and weaknesses.
The big plus is that the real data is available to anybody with the key – for example
administration personnel might be able to see the personal details on their front end
screens but no one else would have this capability. This “optional” visibility is also
this techniques biggest weakness. The encryption password only needs to escape once
and all of t |
he data is compromised. Of course, you can change the key and regenerate
the test instances – but stored or saved copies of the data are immediately available
under the old password.
Encryption also destroys the formatting and look and feel of the data. Encrypted data
rarely looks meaningful, in fact, it usually looks like binary data. This sometimes
leads to NLS character set issues when manipulating encrypted varchar fields. Certain
types of encryption impose constraints on the data format as well |
. For example, the
Oracle Obfuscation toolkit requires that all data to be encrypted should have a length
which is a multiple of 8 characters. In effect, this means that the fields must be
extended with a suitable padding character which must then be stripped off at
decryption time.
The strength of the encryption is also an issue. Some encryption is more secure than
others. According to the experts, most encryption systems can be broken – it is just a
matter of time and effort. In other words, not |
very much will keep the national
security agencies of largish countries from reading your files should they choose to do
so. This may not be a big worry if the requirement is to protect proprietary business
information.
Verdict: The security is dependent on the strength of the encryption used. It may not
be suitable for high security requirements or where the encryption key cannot be
secured. Encryption also destroys the look and feel of the sanitized data. The big plus
is the selective access it p |
resents.
Data Sanitization Techniques
A Net 2000 Ltd. White Paper
Copyright © Net 2000 Ltd. 2003-2004
http://www.Net2000Ltd.com
Summary
Given the legal and business operating environment of today, most test and
development databases will require some form of Data Sanitization. There are a
variety of techniques available and usually several will be required as the format, size
and structure of the data dictates.
One key issue not dis |
cussed above is repeatability. When designing the Data
Sanitization routines it should be realized that they will eventually become a
production process – even if the data is only destined for test environments. In other
words, the data will need to be sanitized each and every time a test database is
refreshed from production. This means that Data Sanitization routines that are easy to
run and simple to maintain will soon recover any extra development effort or costs.
About the Author
Dale Edga |
r in his many years of experience as a DBA has built numerous test and
development environments. He is one of the creative influences behind the Data
Masker - software which provides an automated solution to the sanitization of data in
test and development environments. Dale can be reached at
[email protected]
|
Title:
A Framework of Principles for the
Development of Policies, Strategies and
Standards for the Long-term Preservation of
Digital Records
Status:
Final (public)
Version:
1.2
Submission Date:
June 2005
Release Date:
March 2008
Author:
The InterPARES 2 Project
Writer(s):
Luciana Duranti, Jim Suderman and Malcolm Todd
Project Unit:
Policy Cross-domain
URL:
http://www.interpares.org/display_file.cfm?doc=
ip2(pub)policy_framework_document.pdf
Policy Frame |
work, v1.2 (March 2008)
L. Duranti, J. Suderman and M. Todd
Table of Contents
INTRODUCTION ...............................................................................................................................
1
STRUCTURE OF THE PRINCIPLES
.....................................................................................................
3
PRINCIPLES FOR RECORDS CREATORS ...........................................................................................
4
(C1)
Digital objects m |
ust have a stable content and a fixed documentary form to be considered records
and to be capable of being preserved over time. (P5) .....................................................................................
4
(C2)
Record creation procedures should ensure that digital components of records can be separately
maintained and reassembled over time. (P4) ..................................................................................................
5
(C3)
Record creation and maintenance r |
equirements should be formulated in terms of the purposes the
records are to fulfil, rather than in terms of the available or chosen record-making or recordkeeping
technologies. (P6)
............................................................................................................................................
5
(C4)
Record creation and maintenance policies, strategies and standards should address the issues of
record reliability, accuracy and authenticity expressly and separately. (P2) |
....................................................
6
(C5)
A trusted record-making system should be used to generate records that can be presumed reliable. ............
7
(C6)
A trusted recordkeeping system should be used to maintain records that can be presumed accurate
and authentic. (P11, P12) ................................................................................................................................
8
(C7)
Preservation considerations should be embedded in all activities |
involved in record creation and
maintenance if a creator wishes to maintain and preserve accurate and authentic records beyond its
operational business needs. (P7) ....................................................................................................................
9
(C8)
A trusted custodian should be designated as the preserver of the creator’s records. (P1) ..............................
9
(C9)
All business processes that contribute to the creation and/or use of the same records sho |
uld be
explicitly documented. (P10)
..........................................................................................................................
10
(C10)
Third-party intellectual property rights attached to the creator’s records should be explicitly identified
and managed in the record-making and recordkeeping systems. (P8)
..........................................................
11
(C11)
Privacy rights and obligations attached to the creator’s records should be explicitly identi |
fied and
protected in the record-making and recordkeeping systems. (P9)
.................................................................
11
(C12)
Procedures for sharing records across different jurisdictions should be established on the basis of
the legal requirements under which the records are created. (P13) ..............................................................
12
(C13)
Reproductions of a record made by the creator in its usual and ordinary course of business and for
its purposes an |
d use, as part of its recordkeeping activities, have the same effects as the first
manifestation, and each is to be considered at any given time the record of the creator. (P3)
......................
12
PRINCIPLES FOR RECORDS PRESERVERS .....................................................................................
13
(P1)
A designated records preserver fulfils the role of trusted custodian. (C8)
......................................................
13
(P2)
Records preservation policies, |
strategies and standards should address the issues of record
accuracy and authenticity expressly and separately. (C4)
.............................................................................
14
(P3)
Reproductions of a creator’s records made for purposes of preservation by their trusted custodian
are to be considered authentic copies of the creator’s records. (C13)
...........................................................
15
(P4)
Records preservation procedures should ensure that the digit |
al components of records can be
separately preserved and reassembled over time. (C2) ................................................................................
15
(P5)
Authentic copies should be made for preservation purposes only from the creator’s records; that is,
from digital objects that have a stable content and a fixed documentary form. (C1)
......................................
16
(P6)
Preservation requirements should be articulated in terms of the purpose or desired outcome of
p |
reservation, rather than in terms of the specific technologies available. (C3) ..............................................
17
(P7)
Preservation considerations should be embedded in all activities involved in each phase of the
records lifecycle if their continuing authentic existence over the long term is to be ensured. (C7) ................
18
(P8)
Third-party intellectual property rights attached to the creator’s records should be explicitly identified
and managed in the preservation system. ( |
C10) ...........................................................................................
19
(P9)
Privacy rights and obligations attached to the creator’s records should be explicitly identified and
protected in the preservation system. (C11) ..................................................................................................
19
(P10)
Archival appraisal should identify and analyze all the business processes that contribute to the
creation and/or use of the same records. (C9)
|
...............................................................................................
20
(P11)
Archival appraisal should assess the authenticity of the records. (C6) ..........................................................
20
(P12)
Archival description should be used as a collective authentication of the records in an archival
fonds. (C6) .....................................................................................................................................................
|
20
(P13)
Procedures for providing access to records created in one jurisdiction to users in other jurisdictions
should be established on the basis of the legal environment in which the records were created. (C13) .......
21
InterPARES 2 Project, Policy Cross-domain
i
Policy Framework, v1.2 (March 2008)
L. Duranti, J. Suderman and M. Todd
InterPARES 2 Project, Policy Cross-domain
Page 1 of 1
A Framework of Principles for the Development of Policies, Strategies
and Standards for the Long-te |
rm Preservation of Digital Records1
Introduction
The InterPARES research projects have examined the creation, maintenance and
preservation of digital records. A major finding of the research is that, to preserve trustworthy
digital records (i.e., records that can be demonstrated to be reliable, accurate and authentic),
records creators must create them in such a way that it is possible to maintain and preserve
them. This entails that a relationship between a records creator2 and its designated preserv |
er3
must begin at the time the records are created.4
The InterPARES 1 research (1999-2001) was undertaken from the viewpoint of the
preserver. Three central findings emerged from it: 1) there are several requirements that should
be in place in any recordkeeping environment aiming to create reliable and accurate digital
records and to maintain authentic records;5 2) it is not possible to preserve digital records but
only the ability to reproduce them;6 and 3) the preserver needs to be involved with th |
e records
from the beginning of their lifecycle to be able to assert that the copies that will be selected for
permanent preservation are indeed authentic copies of the creator’s records.
The InterPARES 2 research (2002-2006) took the records creator’s perspective. The
researchers carried out case studies of records creation and maintenance in the arts, sciences
and e-government; they modeled the many functions that make up records creation and
maintenance and records preservation according to both t |
he lifecycle and the continuum
models; they reviewed and compared legislation and government policies from a number of
different countries and at different levels of government, from the national to the municipal; they
analyzed many metadata initiatives and developed a tool to identify the strengths and
weaknesses of existing metadata schemas in relation to questions of reliability, accuracy and
authenticity; and, once again, they studied the concept of trustworthiness and its components,
reliability, |
accuracy and authenticity and how it is understood, not just in the traditional legal and
administrative environments, but in the arts, in the sciences and in the developing areas of e-
government.
1 The term initially used in the InterPARES Project is “electronic records.” In fact, the book resulting from InterPARES 1 is named
The Long-term Preservation of Authentic Electronic Records: Findings of the InterPARES Project (Luciana Duranti, ed.; San
Mini |
ato, Archilab, 2005), and the formal title of InterPARES 2 carries that terminology forward. However, in the course of the
research, the term “electronic record” began to be gradually replaced by the term “digital record,” which has a less generic meaning,
and by the end of the research cycle, the research team had developed separate definitions for the two terms and decided to use
the latter as the one that better describes the object of InterPARES research. The definition for “electronic record” reads: |
“An
analogue or digital record that is carried by an electrical conductor and requires the use of equipment to be intelligible by a person.”
The definition for “digital record” reads: “A record whose content and form are encoded using discrete numeric values (such as the
binary values 0 and 1) rather than a continuous spectrum of values (such as those generated by an analogue system).” See the
InterPARES 2 Terminology Database, available at http://www.interpares.org/ip2/ip2_terminology_db.cfm.
2 Recor |
ds creator is the physical or juridical person (i.e., a collection or succession of physical persons, such as an organization, a
committee, or a position) who makes or receives and sets aside the records for action or reference. As such, the term includes all
officers who work for a juridical person, such as records managers, records keepers and preservers.
3 Records preserver is a generic term that refers more to the function than to the professional designation of the physical or juridical
person in q |
uestion. Thus, the preserver might be a unit in an organization, a stand-alone institution, an archivist or anyone else who
has as primary responsibility the long-term preservation of records.
4 Records are created when they are made or received and set aside or saved for action or reference.
5 See Authenticity Task Force (2002). “Appendix 2: Requirements for Assessing and Maintaining the Authenticity of Electronic
Records,” in The Long-term Preservation of Authentic Electronic Records: Findings of the |
InterPARES Project, Luciana Duranti, ed.
(San Miniato, Italy: Archilab, 2005), 204–219. PDF version available at
http://www.interpares.org/book/interpares_book_k_app02.pdf.
6 See Kenneth Thibodeau et al., “Part Three – Trusting to Time: Preserving Authentic Records in the Long Term: Preservation Task
Force Report,” ibid, 99–116. PDF version available at http://www.interpares.org/book/interpares_book_f_part3.pdf.
Policy Framework, v1.2 (March 2008)
L. Duranti, J. Suderman and M. Todd
InterPARES 2 Proj |
ect, Policy Cross-domain
Page 2 of 2
The case studies showed that record creation in the digital environment is almost never
guided by considerations of preservation over the long term. As a result, the reliability, accuracy
and authenticity of digital records can either not be established in the first place or not be
demonstrated over periods of time relevant to the “business”7 requirements for the records.
These records cannot therefore support the creator’s accountability requirements, nor can th |
ey
be effectively relied upon either by the creator for reference or later action or by external users
as sources. Furthermore, they cannot be understood within an historical context, thereby
undermining the traditional role of preserving organizations such as public archival institutions.
The research undertaken in records and information-related legislation showed that no level
of government in any country to date has taken a comprehensive view of the records lifecycle,
and that, in some cases, leg |
islation has established significant barriers to the effective
preservation of digital records over the long term, most notably that regarding copyright.
It was the responsibility of the InterPARES 2 Policy Cross-domain research team
(hereinafter “the Policy team”) to determine whether it was possible to establish a framework of
principles that could guide the creation of policies, strategies and standards, and that would be
flexible enough to be useful in differing national environments, and consiste |
nt enough to be
adopted in its entirety as a solid basis for any such document. In particular, such a framework
had to balance different cultural, social and juridical perspectives on the issues of access to
information, data privacy and intellectual property.
The findings of the InterPARES 1 research were confirmed by the research conducted by
the InterPARES 2 Policy team, which further concluded that it is possible to develop such a
framework of principles to support record creation, maintenance an |
d preservation, regardless of
jurisdiction. This document, in combination with other products of the Project, especially the
“Chain of Preservation model,”8 reflects this conclusion, while emphasizing the need to make
explicit the nature of the relationship between records creators and preservers.
The Policy team developed two complementary sets of principles, one for records creators
and one for records preservers, which are intended to support the establishment of the
relationship between creators |
and preservers by demonstrating the nature of that relationship.9
The principles for records creators are directed to the persons responsible for developing
policies and strategies for the creation, maintenance and use of digital records within any kind of
organization, and to national and international standards bodies. The principles for records
preservers are directed to the persons responsible for developing policies and strategies for the
long-term preservation of digital records within administra |
tive units or institutions that have as
their core mandate the preservation of the bodies of records created by persons, administrative
units or organizations external to them, selected for permanent preservation under their
jurisdiction for reasons of legal, administrative or historical accountability. They are therefore
intended for administrative units (e.g., a bank, a city or a university archives) or institutions (e.g.,
a community archives or a state archives) with effective knowledge of records |
and records
preservation.
7 The term “business” is used in its most general sense, since the object of the InterPARES research includes works of art and
scientific data as well as standard types of business records.
8 The model is available at http://www.interpares.org/ip2/ip2_models.cfm.
9 The initial draft of the principles relied heavily on the contributions of three research assistants: Fiorella Foscarini, Emily O’Neill and
Sherry Xie.
Policy Fram |
ework, v1.2 (March 2008)
L. Duranti, J. Suderman and M. Todd
Structure of the Principles
The principles are similarly presented, with the principle statement followed by an
explanatory narrative, sometimes with illustrative examples. The principles are more often
phrased as recommendations (“should”) rather than imperatives (“must”), because some of
them might not be relevant to some records creators or preservers. Each principle statement is
followed by an indication of the corresponding principle i |
n the other set (C stands for Creator, P
stands for Preserver; the number is the principle number in the C or the P set). The reason why
the principle numbers do not correspond in the two sets (C1=P1) is that the principles are listed
in each set in order of relative importance.
InterPARES 2 Project, Policy Cross-domain
Page 3 of 3
Policy Framework, v1.2 (March 2008)
L. Duranti, J. Suderman and M. Todd
InterPARES 2 Project, Policy Cross-domain
Page 4 of 4
Principles for Records Creators
(C1) |
Digital objects must have a stable content and a fixed documentary form to
be considered records and to be capable of being preserved over time. (P5)
The InterPARES Project has defined a record as “a document made or received in the
course of a practical activity as an instrument or a by-product of such activity, and set aside for
action or reference,”10 adopting the traditional archival definition. This definition implies that, to
be considered as a record, a digital object generated by the creator mu |
st first be a document;
that is, must have stable content and fixed documentary form. Only digital objects possessing
both are capable of serving the record’s memorial function.
The concept of stable content is self-explanatory, as it simply refers to the fact that the data
and the information in the record (i.e., the message the record is intended to convey) are
unchanged and unchangeable. This implies that data or information cannot be overwritten,
altered, deleted or added to. Thus, if one has a s |
ystem that contains fluid, ever-changing data or
information, one has no records in such a system until one decides to make one and to save it
with its unalterable content.
The concept of fixed form is more complex. A digital object has a fixed form when its binary
content is stored so that the message it conveys can be rendered with the same documentary
presentation it had on the screen when first saved. Because the same documentary
presentation of a record can be produced by a variety of digital fo |
rmats or presentations,11 fixed
form does not imply that the bitstreams must remain intact over time. It is possible to change the
way a record is contained in a computer file without changing the record; for example, if a digital
object generated in ‘.doc’ format is later saved in ‘.pdf’ format, the way it manifests itself on the
screen—its documentary presentation, or “documentary form”—has not changed, so one can
say that the object has a fixed form.
One can also produce digital information that c |
an take several different documentary forms.
This means that the same content can be presented on the screen in several different ways, the
various types of graphs available in spreadsheet software being one example. In this case, each
presentation of such a digital object in the limited series of possibilities allowed by the system is
to be considered as a different view of the same record having stable content and fixed form.
In addition, one has to consider the concept of “bounded variability,” whic |
h refers to changes
to the form and/or content of a digital record that are limited and controlled by fixed rules, so that
the same query, request or interaction always generates the same result.12 In such cases,
variations in the record’s form and content are either caused by technology, such as different
operating systems or applications used to access the document, or by the intention of the author
or writer of the document. Where content is concerned, the same query will always return the
same sub |
set, while, as mentioned, its presentation might vary within an allowed range, such as
image magnification. In consideration of the fact that what causes these variations also limits
them, they are not considered to be violations of the requirements of stable content and fixed
form.
10 See InterPARES 2 Terminology Database, op. cit.
11 Digital format is defined as “The byte-serialized encoding of a digital object that defines the syntactic and semantic |
rules for the
mapping from an information model to a byte stream and the inverse mapping from that byte stream back to the original information
model” (InterPARES 2 Terminology Database, op. cit.). In most contexts, digital format is used interchangeably with digital file-
related concepts such as file format, file wrapper, file encoding, etc. However, there are some contexts, “such as the network
transport of formatted content streams or consideration of content streams at a level of granularity finer |
than that of an entire file,
where specific reference to “file” is inappropriate” (Stephen L. Abrams (2005), “Establishing a Global Digital Format Registry,”
Library Trends 54(1): 126. Available at http://muse.jhu.edu/demo/library_trends/v054/54.1abrams.pdf).
12 See Luciana Duranti and Kenneth Thibodeau (2006). “The Concept of Record in Interactive, Experiential and Dynamic
Environments: the View of InterPARES,” Archival Science 6(1): 13-68.
Policy Framework, v1.2 (March 2008)
L. Duranti, J. Suderman |
and M. Todd
Organizations should establish criteria for determining which digital objects need to be
maintained as records and what methods should be employed to fix their form and content if
they are fluid when generated. The criteria should be based on business needs but should
respect as well the requirements of legal, administrative and historical accountability.
(C2) Record creation procedures should ensure that digital components of
records can be separately maintained and reassembled over time. |
(P4)
Every digital record is composed of one or more digital components. A digital component is
a digital object that is part of one or more digital records, including any metadata necessary to
order, structure or manifest content, and that requires a given preservation action. For example,
an e-mail that includes a picture and a digital signature will have at least four digital components
(the header, the text, the picture and the digital signature). Reports with attachments in different
formats wil |
l consist of more than one digital component, whereas a report with its attachments
saved in one PDF file will consist of only one digital component. Although digital components
are each stored separately, each digital component exists in a specific relationship to the other
digital components that make up the record.
Preservation of digital records requires that all the digital components of a record be
consistently identified, linked and stored in a way that they can be retrieved and reconstituted
|
into a record having the same documentary presentation it manifested when last closed. Each
digital component requires one or more specific methods for decoding the bitstream and for
presenting it for use over time. The bitstream can be altered, as a result of conversion for
example, as long as it continues to be able to fulfil its original role in the reproduction of the
record. All digital components must be able to work together after they are altered; therefore, all
changes need to be assessed by t |
he creator for the effects they may have on the record.
Organizations should establish policies and procedures that stipulate the identification of
digital components at the creation stage and that ensure they can be maintained, transmitted,
reproduced, upgraded and reassembled over time.
(C3) Record creation and maintenance requirements should be formulated in
terms of the purposes the records are to fulfil, rather than in terms of the
available or chosen record-making or recordkeeping technologies. |
(P6)
Digital records rely, by definition, on computer technology and any instance of a record
exists within a specific technological environment. For this reason, it may seem useful to
establish record creation and maintenance requirements in terms of the technological
characteristics of the records or the technological applications in which the records may reside.
However, not only do technologies change, sometimes very frequently, but they are also
governed by proprietary considerations established |
and modified at will by their developers.
Both these factors can significantly affect the accessibility of records over time. For these
reasons, references to specific technologies should not be included in records policies,
strategies and standards governing the creation and maintenance of an organization’s records.
Only the business requirements and obligations that the records are designed to support should
be explicitly kept in consideration at such a high regulatory level. At the level of impleme |
ntation,
the characteristics of specific technologies should be taken into account to support the
established business requirement and make possible its realization.
Technological solutions to record creation and maintenance are dynamic, meaning that they
will evolve as the technology evolves. New technologies will enable new ways of creating
records that meet an organization’s business requirements. The rapid adoption of Web
technologies to support business communication and transaction illustrates |
this. Specific
activities for maintaining records will therefore require continuing adaptation to new situations
InterPARES 2 Project, Policy Cross-domain
Page 5 of 5
Policy Framework, v1.2 (March 2008)
L. Duranti, J. Suderman and M. Todd
drawing on expertise from a number of disciplines. To extend the example of the use of Web
technologies, organizations creating and maintaining transactional records in a mainframe
environment need to draw on knowledge of the new Web technologies from both connec |
tivity
(i.e., how to connect the mainframe to the Web) and security standpoints (i.e., how to protect
the records from remote, Web-based attacks). As new technologies are used to create records,
reference to new archival knowledge will continue to be required.
Technological solutions need to be specific to be effective. Although the general theory and
methodology of digital preservation applies to all digital records, the maintenance solutions for
different types of records require different methods. |
Therefore, they should be based on the
specific juridical-administrative context in which the records are created and maintained, the
mandate, mission or goals of their creator, the functions and activities in which the records
participate and the technologies employed in their creation to ensure the best solutions are
adopted for their maintenance.
Record policies that are expressed in terms of business requirements rather than
technologies will need to be periodically updated as the organization’s |
business requirements
change, rather than as the technology changes. It is the role of a specific action plan to identify
appropriate technological solutions for the maintenance of specific aggregations of records. The
identified solutions must be monitored with regard to the possible need for modifying and
updating. This requires the records creating body to be aware of new research developments in
the archival and records management fields and to collaborate with interdisciplinary efforts to
develo |
p appropriate methods for the management of digital records.
(C4) Record creation and maintenance policies, strategies and standards should
address the issues of record reliability, accuracy and authenticity expressly and
separately. (P2)
In the management of digital records, reliability, accuracy and authenticity are three vital
considerations for any organization that wishes to sustain its business competitiveness and to
comply with legislative and regulatory requirements. These considerations shoul |
d be directly
and separately addressed in records policies and promulgated throughout the organization. The
concept of reliability refers to the authority and trustworthiness of a record as a representation of
the fact(s) it is about; that is, to its ability to stand for what it speaks of. In other words, reliability
is the trustworthiness of a record’s content. It can be inferred from two things: the degree of
completeness of a record’s documentary form and the degree of control exercised over the
pr |
ocedure (or workflow) in the course of which the record is generated. Reliability is then
exclusively linked to a record’s authorship and is the sole responsibility of the individual or
organization that makes the record. Because, by definition, the content of a reliable record is
trustworthy, and trustworthy content is, in turn, predicated on accurate data, it follows that a
reliable record is also an accurate record.
An accurate record is one that contains correct, precise and exact data. Accuracy of |
a
record may also indicate the absoluteness of the data it reports or its perfect or exclusive
pertinence to the matter in question. The accuracy of a record is assumed when the record is
created and used in the course of business processes to carry out business functions, based on
the assumption that inaccurate records harm business interests. However, when records are
transmitted across systems, refreshed, converted or migrated for continuous use, or the
technology in which the record resides is up |
graded, the data contained in the record must be
verified to ensure their accuracy was not harmed by technical or human errors occurring in the
transmission or transformation processes. The accuracy of the data must also be verified when
records are created by importing data from other records systems. This verification of accuracy
is the responsibility of the physical or juridical person receiving the data; however, such person
is not responsible for the correctness of the data value, for which the se |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.