EN - Data Sanitization and Recovery
EN - Data Sanitization and Recovery
EN - Data Sanitization and Recovery
Summary
Summary: user data is left on disk drives removed from computers and storage systems,
creating a data security vulnerability that many users are unaware of. Recent Federal and
state laws requiring secure erasure of user data expose companies to fines of $250,000
and responsible parties to imprisonment for 10 years.
Complete eradication of user data off drives can be accomplished by running data Secure
Erasure utilities such as the freeware “HDDerase” downloadable here. It executes the
Federally-approved (NIST 800-88) Secure Erase command in the ATA ANSI standard,
which is implemented in all recent ATA drives greater than 15-20 GB. A similar
command in the SCSI ANSI standard is optional and not yet implemented in drives
tested. Normal Secure Erase takes 30-60 minutes to complete. Some ATA drives also
implement the standard Enhanced Secure Erase command that takes only milliseconds to
complete.
Table of Contents
Introduction
Data security has risen to be one of the highest concerns of computer professionals.
Tighter legal requirements now exist for protecting user data from unauthorized use, and
1
for both preserving and erasing (sanitizing) records to meet legal compliance
requirements. This Tutorial document will address concerns and developments in the
sanitization and protection of user data.
Overall data storage security entails protection at different levels and locations:
• Data at rest - drive data erasure
• Secure erase of all data blocks on disk drives
• Single file erasure
• Drive physical or magnetic destruction
• Data in motion - data encrypted during transport
• Protection of data and crypto keys during transport
• Transparency to users (automatic encryption)
• Drive internal encryption (data encrypted by storage device)
• Access level dependent upon key or password used to decrypt data
• Drive data sanitization
• Secure erasure of user data for drive disposal or reuse
The following table (Table 1) outlines comparative times to execute various approaches
for data sanitization (erasure) as well as level of data sanitization security.
DoD 5220 Block Up to several days Medium Need 3 writes + verify, cannot erase
Erase reassigned blocks
Secure Erase 1-2 hours High In-drive overwrite of all user accessible
records
NIST 800-88 Seconds Very high Change in-drive encryption key
Enhanced Secure
Erase
2
All this means that true computer data erasure is an abnormal event. These measures
taken to protect and speed access to user data can make that data vulnerable to recovery
by unauthorized persons.
When a computer is lost or disposed of, active and discarded data typically remains
stored on its hard disk drive. Even if users “delete” all their files, they can be recovered
from “recycling” folders or by special utility programs such as Norton Unerase.
If data is not erased beyond recovery, data on disk drives that leave the physical control
of owners can and often does fall into the hands of others. Data can be recovered with
little effort, from discarded, warranty repaired, or resold disk drives. Many reports have
been written on data recovered from discarded disk drives 1,2. Each year hundreds of
thousands of hard disk drives are retired. Some of these hard disk drives find their way
back into the market and their data can be recovered unless it is erased securely.
There is an urgent need for a capability to reliably erase data and prevent access to data
from retired computer hard disk drives for security and privacy reasons. Data sanitization
needs arise differently depending upon the user application. Even consumer drives could
use data sanitization to protect user privacy or for DRM purposes.
There are several laws and regulations that relate to data retention and data sanitization
on data storage devices like hard disk drives. Some US requirements are listed below:
3
Gramm-Leach-Bliley Act (GLBA)
California Senate Bill 1386
Sarbanes-Oxley Act (SBA)
SEC Rule 17a
The Federal Health Insurance Portability and Accountability Act (HIPAA) sets goals on
keeping personal information secure in the health industry. If a company is found in
non-compliance of HIPAA data security practices, the company may be exposed to a
maximum fine of $250,000 and the responsible party can face a maximum of 10 years
imprisonment.
There are several approved methods for data sanitization that satisfy these legal
requirements or meet even more stringent corporate or government secrecy requirements.
Many of them physically destroy disk drives to prevent any future use. Another data
security measure is encryption of user data.. Secure data encryption from creation to
destruction is approved by some regulatory compliance legislation to protect sensitive
information. Its security level is determined by Federal document FIPS 142-2.
Secure erase is recognized by NIST 800-88 as an effective and secure way to meet legal
data sanitization requirements against attacks up to laboratory level..
3
NIST Special Publication 800-88, Guidelines for Media Sanitization, August 2006
4
Legal Penalties for Failure to Sanitize Data4
The following table summarizes the fines and jail penalties for violation of the data
security laws.
Gramm-Leach-
Sarbanes-Oxley FACTA HIPAA
Bliley
Public Company
Fair and Accurate Health Insurance
Financial Services Accounting Reform
Credit Transaction Portability &
Modernization Act & Investor
Act Accountability Act
Protection Act
Directors and
$10,000 $1,000,000 $50,000 to $250,000
Officers
Institution $100,000
Years in Prison 5 to 12 years 20 years 1 to 10 years
FDIC Insurance Terminated
Impact on
Cease and Desist
Operations
Individual $1,000,000 Civil Action $25,000
Institution 1% of assets
Four basic sanitization security levels can be defined: weak erase (deleting files), block
erase (overwrite by external software), normal secure erase (current drives), and
enhanced secure erase (see below). UCSD’s CMRR has established test protocols for
software secure erase5.
Block erase is most commonly used. While it significantly better than no erase, or file
deletion, or drive formatting, it is vulnerable to malware and incomplete erasure of all
data blocks. Examples are data blocks reassigned by drives, multiple drive partitions, host
protected areas, device configuration overlays, and drive faults.
Normal secure erase is approved by NIST 800-88 for legal sanitization of user data up to
Confidential, and enhanced secure erase for higher levels. Enhanced level has only
recently been implemented, initially in Seagate drives, and these drives are under
evaluation by the CMRR.
These four erasure protocols exist because users make tradeoffs between sanitization
security level and the time required. A high security protocol that requires special
software and days to accomplish will be avoided by most users, making it little used and
of limited practical value. For example, the old data overwrite document DoD 5220 calls
for multiple block overwrites of Confidential data, which can take more than a day to
4
From Ensconce Data Technology, Inc
5
G. Hughes, CMRR Protocols for Disk Drive Secure Erase, cmrr.ucsd.edu/Hughes/
5
complete in today’s large capacity drives. So users make tradeoffs between the time
required to erase data and the risk that the next drive user may know and use recovery
techniques which can access weakly erased data. Figure 1 shows tradeoffs in security
level vs. speed of erasure for various erasure options.
Figure 1. Security vs. Speed of Completion of Various Modes to Erase Data on Hard
Disk Drives
For all but top-secret information, and when time is important, users will often turn to
erasure that takes minutes rather than hours or days. They will select a method giving
them an acceptable level of security in a reasonable time window.
Some storage products are more easily destroyed than hard disk drives, including
magnetic disk data cartridges, tape cartridges, secure USB drives, and optical media.
6
The CMRR studies the adequacy of degaussers for data sanitization.
One issue with disk drive degaussing is that it is normally done on intact disk drives. AC
field degaussing is commonly used, and conductive components in the hard disk drive
can cause some shielding of the disks from the magnetic fields. As a consequence, lower
frequency and higher intensity magnetic fields are required than would be the case if the
disks were removed from the disk drives and exposed directly to the magnetic fields.
Drive designers continually increase the linear density of magnetic recording to create
higher data storage capacity per disk. This raises the disk magnetic coercivity, the field
required to write bits on the magnetic media. As the magnetic coercivity increases, the
fields required to erase the data on recorded disks increases. Thus an older degausser may
not fully erase data on a newer hard disk drive. New perpendicular recording drives may
not be erasable by present degaussers designed for past longitudinal recording drives. As
the degausser field increases the cost and power usage increases as well.
Future generations of magnetic recording media may use very high magnetic coercivity
disks to achieve areal densities greater than 500 gigabits per square inch. These drives
may have technology using laser light in the magnetic write element of the disk drive, to
raise the temperature of a spot on the magnetic medium in order to lower the magnetic
coercivity to the point where the write element can record a bit on the very high
coercivity magnetic media. For disk drives using this Heat or Thermally Assisted
Magnetic Recording (HAMR/TAMR) technology the degausser field required to erase
the disk drive at room temperatures may be impossible or impractical to achieve. In this
case the drive may have to be physically destroyed.
“Hybrid drives” are now being introduced for notebook or laptop computers that have
flash memory write cache on hard disk drive circuit boards. Magnetic degaussing would
not affect any resident data on such semiconductor memory chips. Data on these
non-volatile semiconductors would have to be sanitized using some other technique. For
all these reasons degaussing of all the data on hard disk drives will become increasingly
impractical.
7
It is difficult for external software to reliably sanitize user data stored on a hard disk
drive. Many commercial software packages are available using variations of DoD 5220,
making as many as 35 overwrite passes. But in today’s drives, multiple overwrites are no
more effective than a single overwrite. Off-track overwrites could be effective in some
drives, but there is no such drive external command for a software utility to move heads
offtrack. And even three overwrites can take more than a day to erase a large capacity
hard disk drive. In busy IT facilities, such time is often not available and IT personnel are
likely to take short cuts.
DoD 5220 overwriting has other vulnerabilities, such as erasing only to a drive’s
Maximum Address which can be set lower than its native capacity; not erasing
reallocated (error) blocks; or miss extra partitions. External overwrites cannot access the
reallocated sectors on most drives, and any data once recorded is left on these sectors.
These sectors could conceivably be recovered and decoded by exotic forensics. While
enterprise-class drives and drive systems (SCSI/FC/SAS/iSCSI) allow software
commands to test all the user blocks for write and read ability, mass market drives
(PATA/SATA) cannot read, write or detect reassigned blocks since they have no logical
block address.
The Secure Erase (SE) command was added to the open ANSI standards that control disk
drives, at the request of CMRR at UCSD. The ANSI T13.org committee oversees the
ATA interface specification (also called IDE) and the ANSI T10.org committee governs
the SCSI interface specification.
Secure erase is built into the hard disk drive itself and thus is far less susceptible to
malicious software attack than external software utilities.
The SE command is implemented in all ATA interface drives manufactured after 2001
(drives with capacities greater than 15 GB), according to testing by CMRR. A
standardized internal secure erase command also exists for SCSI drives, but is optional
and not currently implemented in SCSI drives tested.
Secure erase does a single on-track erasure of the data on the disk drive. The U.S.
National Security Agency published an Information Assurance Approval of single pass
8
overwrite, after technical testing at CMRR showed that multiple on-track overwrite
passes gave no additional erasure.
Secure erase has been approved by the U.S. National Institute for Standards and
Technology (NIST), Computer Security Resource Center 6. NIST document 800-88
approves SE at a higher security level than external software block overwrite utilities like
as Norton Government Wipe. Software overwrite utilities are approved by NIST 800-88
for lower security sanitization, but they may not meet the legal requirements of HIPAA,
PIPEDA, GLBA, or Sarbanes-Oxley.
Drive manufacturers today are pursuing higher security secure erase (including secret
data), via in-drive data encryption (see below)
Why encrypt data at rest in drives instead of in computers, such as by user application
programs that access the data? Because computer level data encryption defeats the
purpose of many important data management functions, including incremental backup,
continuous data protection, data compression, de-duplication, virtualization, archiving,
content addressable storage, advanced routing, and thin provisioning9. Defeating these
operations causes significant penalties in data access speed and cost, to enterprise storage
companies. Each of these operations exploits the structure of user data, and need to
inspect the data in order to that inspect data function. They become inefficient or
nonfunctional if the data has been randomized by encryption. For example, data
compression ratios may fall from more than 2:1 to less than 1:1, because compressing
random data can expand it instead. De-duplication won’t find identical data sets if they
are encrypted by different users. Computer level encryption can be employed with
in-drive encryption as well, the double encryption does no harm and gives additional
security. In-drive encryption b can relieve encryption key management problems inherent
6
NIST Computer Security Resource Center, Special Publication 800-88: Guidelines for Media
Sanitization, August 2006
7
G. Hughes, “Wise Drives”, IEEE Spectrum, August 2002
8
e.g. Seagate Momentus 5200 drives
9
Storage magazine, October 2006
9
in removable storage, like laptop disk drives or tape backups. In fact, hardware-based
tape drive encryption may become widespread10 by 2007 due to widely publicized losses
of backup tape reels containing identity theft data on millions of people.
Full Disk Encryption (FDE) Enhanced Secure Erase,” (“FDE-SE”), securely changes the
internal drive encryption key, to render encrypted user data on disk indecipherable. This
is enabled via the Enhanced SE command in the existing ATA ANSI specs.
FDE SE encryption needs to be tested for protection against advanced forensic analysis.
The results will determine the erasure security data level - Confidential, Secret, or higher.
The US Commerce Department prohibits most 256-bit and higher encryption export
overseas, limiting FDE E-SE to AES-128-bit encryption, since disk drives are a global
industry.
AES-256 bit encryption in FDE drives could allow FDE SE at a higher security level.
However, an FDE E-SE operation amounts to double AES-128, because the data
encrypted by the discarded key is decrypted by the new key, and AES is a symmetric
encryption scheme. It would appear that a brute force attack on double AES-128 requires
the same computational effort as single AES-256.
For the highest security allowed by NIST 800-88, the cypt-text in an FDE disk drive
could be eliminated by a Normal OW SE done after the FDE E-SE.
An open industry standard for FDE is being worked on by the Trusted Computing Group
overall specification (the Storage Working Group in trustedcomputinggroup.org). Drive
members of the TCG include Seagate, HGST, Fujitsu and WD. SE via encryption may be
included, consistent with the ANSI open standards for ATA drives (t13.org)
CMRR has begun testing FDE-SE drives. They take less than 15 milliseconds to
complete an Enhanced SE - while a 750 GB ATA-interface HDD can take over an hour
to erase using conventional Secure Erase (or many hours using external overwrite
software).
10
without using normal drive electronics 11. Reading back tracks from a disk taken out of a
drive and tested on a spin stand was practical decades ago, but no longer with today’s
microinch-size tracks.
The time required by exotic technologies is itself a barrier to data recovery and increases
data security. Also, accessing data from magnetic images requires overcoming almost a
dozen successive magnetic recording technology hurdles. Even if these hurdles were
overcome, about an hour would be required to recover a single user data block out of
millions on a disk. Recovering substantial amounts of data in less than months requires
that the disk be intact and undamaged, so that heads can be flown over it to obtain data
playback signals; then overcoming these technology hurdles. Simply bending a disk
makes this nearly impossible, so physical damaging drives to warp their disks makes
recovery practically impossible.
Other “experts” claim that limited information can be recovered from unerased track
edges. But this has been shown to be false by tests at CMRR 12. Such recovery also
presumes detailed technical knowledge of the drive’s magnetic recording design. Charles
Sobey at ChannelScience.com wrote an illuminating article on drive-independent data
recovery, showing how difficult these hurdles are.13
Weak deletion by users deleting files in public operating systems such as Windows or
Linux (“usual computer erase’ in Figure 1). This deletes only file directory entries, not
the user data itself.
Block erasure utilities overwrite all user accessible blocks. Block overwriting gives a
higher level of deletion confidence than (1) and these utilities claim to meet Federal
Government requirements in DoD 5220. Today’s hard drive technology has obsoleted
this document, and NIST 800-88 should be used instead.
Disk drive Secure Erase is a drive command defined in the ANSI ATA and SCSI disk
drive interface specifications, which runs inside drive hardware. It completes in about 1/8
the time of 5220 block erasure.
CMRR provides verification and certification of data erasure effectiveness for the
government as well as drive companies and may be the most experienced organization in
the world on disk drive data erasure. It is one of the few public organization with detailed
knowledge of drive internal technology. CMRR requested the SE command now in the
T13.org ATA specification. For Normal Erase mode it requires that the SECURITY
11
www.actionfront.com
12
T. M. Coughlin and G. F. Hughes, “Secure Erase of Disk Drive Data,” IDEMA Insight Magazine, pp.
22-25, Summer 2002
13
See white papers at http://www.actionfront.com/ts_whitepaper.aspx
11
ERASE UNIT command shall write binary zeroes to all user accessible data areas. Note
that ATA reassigned blocks are not user accessible because they have no user address.
CMRR verification testing showed that the erasure security is at the level NIST 800-88,
because drives having the command also randomize user bits before storing on magnetic
media. In-drive block verify is via internal write fault detection hardware, which takes no
additional time thus increases user willingness to use the command. CMRR test times
were up to days for DOD 5220 but the drive normal Secure Erase can complete in 30-45
minutes.
The Department of the Navy licensed secure erase to erase data from disk drives. Some
commercial vendors are also selling products using Secure Erase, such as Esconce Data
Technology.
Coughlin Associates provides data storage consulting and market and technology analysis
of the data storage industry. Visit www.tomcoughlin.com or call 408-871-8808 for more
information.
12
Glossary
ANSI T-10 An ANSI standards committee that overseas the SCSI
interface specification
13
HIPAA Health Information Portability and Accountability Act
Recycle Folder A computer location where “deleted” files are kept until the
recycle folder is emptied
14
array applications. Serial Attached SCSI or SAS is
displacing the older parallel SCSI interfaces.
Secure Erase (SE): A technique for sanitizing all the data stored on a hard disk
drive using internal commands. The data erased can include
reallocated defect sectors
15