System and Method for Encrypted Disk Drive Sanitizing
1. A method of sanitizing a storage device comprising:
- writing a set of known data patterns to predetermined storage locations in the storage device;
issuing a firmware-based erase or wipe command;
reading retrieved data stored at each of the predetermined storage locations;
verifying that the retrieved data does not match the predetermined data pattern.
A system and method for first changing the encryption key on a self-encrypting disk drive followed by a complete disk wipe. Either process can be separately performed, and they can be performed in any order. In fact, one embodiment of the invention, resets the symmetric key, wipes the disk a predetermined number of times with different predetermined data patterns, and then resets the key a second time. This assures that there is absolutely no way to recover the original key or to read the original plain text data, even if some of it'"'"'s encrypted values remain on unallocated tracks after wiping. A user can be assured that in milliseconds after starting the wiping process, the entire disk is rendered unreadable and unrecoverable. Verifiable data can be pre-written to a device that is later read back to assure that wiping or firmware-based erase commands have worked.
- 1. A method of sanitizing a storage device comprising:
writing a set of known data patterns to predetermined storage locations in the storage device; issuing a firmware-based erase or wipe command; reading retrieved data stored at each of the predetermined storage locations; verifying that the retrieved data does not match the predetermined data pattern.
- View Dependent Claims (2, 3, 4)
- 5. A method of assuring sanitization of a storage device comprising performing the following steps in order:
writing a set of known data patterns to predetermined storage locations in the storage device; writing a predetermined wipe data pattern to each address of said disk drive or issuing a firmware-based erase command; reading retrieved data stored at each of the predetermined storage locations; verifying that the retrieved data does not match the predetermined data pattern; issuing a command to said disk drive causing at least one cryptographic key in said disk drive to change value.
- View Dependent Claims (6, 7, 8, 9, 10, 11, 12)
System and Method for Encrypted Disk Drive Sanitizing This is a continuation-in-part of application Ser. No. 15/673,687 filed Aug. 10, 2017. Application Ser. No. 15/673,687 is hereby incorporated by reference in its entirety.
The present invention relates generally to the field of disk drive data sanitation and more particularly to a system and method of automatically resetting an encryption key on an encrypted disk drive before wiping.
Hardware-based full encryption of a disk drive is a technique known in the art to protect data on a disk in a manner that is transparent to any computer or operating system attached to or running the drive. Special circuits on the drive itself perform hardware symmetric encryption of all data being written to the disk and decryption of all data being read from the disk. The encryption is typically a very secure symmetric technique such as the Advanced Encryption Standard (AES) using a 128 or 256 bit key.
Authentication is usually supplied during boot-up as a BIOS event that allows the main key to be made available to the encryption circuitry . The system BIOS may require a password or other security measure. Without this, the drive will not decrypt stored data, and the system cannot boot or otherwise read the hard drive. The drive itself, usually requires an authentication code that may be as long as 32 bits in order to unlock. In most cases, this authentication code is used to decode an encrypted symmetric key using a second encryption technique (that may be identical to the main technique or may be different).
The AES or other key is stored in the hard drive controller hardware or chipset dedicated to encryption. It cannot be read out, and is usually encrypted when not in use. However, it can be changed to a new key with proper authentication. Once the change is made to a new key, none of the data already stored on the disk can be read (more precisely, none of the encrypted, stored data can be correctly decrypted) rendering the disk safe from data loss or readout. A key change is typically accomplished by issuing a cryptographic erase or key reset command to the disk drive after proper authentication. In almost all cases, the disk drive hardware itself generates the new key. That way, the new key is never transferred to or from the outside world in any way, and each new key can be assured to have proper strength (not be a weak key like all ones, all zeros or the like). Some disk drive systems save an encrypted version of the previous symmetric key to allow data recovery in the case of an accidental cryptographic erase; however, this weakens the security and is not widely used. Usually, once a symmetric key is changed, it cannot be recovered by any means. Since it was generated automatically by hardware internal to the disk, it is never transferred outside the disk, and there is no record of it anywhere.
Disk wiping is also known in the art. This is the process of over-writing sectors with known data patterns. It is possible to remove all data from a hard disk thus rendering it safe by writing every sector with a known pattern. Some standards (such as some U.S. Department of Defense standards) require that every sector be over-written three times or some other number of times, and in some cases with a different data pattern each time. This can be very time consuming taking many hours with large disks. Typical disk over-write patterns (on a byte basis) may be alternating patterns such as 0×55 for the first pass, 0×AA for the second pass, and 0×3C for the third pass for example.
There are commercial disk-wiping systems known in the art that can automatically wipe a particular disk according to a particular standard and can be controlled remotely. These systems control the wiping process for start to finish and can guarantee that every sector has been wiped the required number of times.
There are other problems with disk wiping besides the length of time it takes. One of these problems is that some hard disks dynamically allocate tracks to reduce overhead or to remove bad tracks from service. This creates the danger of old plain-text data magnetically (or electronically) residing on a track (or sector) that has been de-allocated and cannot be directly reached by normal addressing. Thus, as a wipe algorithm walks through all possible disk address, it fails to over-write these areas (since they have been dynamically removed from the address space). With some drives, it is possible to locate and wipe these regions; however, with others, there may be no way to be sure that everything has been over-written.
It would be advantageous to have a system and method that could be used with self-encrypting hard disks or other storage devices to quickly make sure that all written data is useless (within milliseconds) no matter where written, and then to also over-write all available sectors a predetermined number of times so that a disk owner is guaranteed there is no useable data left on the device, and that not even the wipe program itself can access any data that was previously stored. It would also be advantageous to have a way of assuring that a wipe has really taken place on the device. This is especially true when using a firmware-based erase command to erase the drive.
The present invention is a system and method for first resetting (changing) the encryption key on a self-encrypting disk drive followed by a complete disk wipe. Either process can be separately performed, and they can be performed in any order. One embodiment of the invention resets the symmetric key, wipes the disk a predetermined number of times with different predetermined data patterns, and then resets the key a second time. This assures that there is absolutely no way to recover the original key or to read the original plain text data, even if some of it'"'"'s encrypted values remain on unallocated tracks after wiping. In addition, a user can be assured that in milliseconds after starting the wiping process, the entire disk is rendered unreadable and unrecoverable. Upon starting the process, the wipe system authenticates itself to the disk controller at a level that allows resetting the main key After changing that key, it then proceeds to wipe every sector the required number of times. Finally, in the double-reset mode, it again changes the key. The disk is safe, clean and in condition for immediate use. Additional formatting can be optionally performed. For any type of wiping process, known data can be pre-written to selected locations before the wipe process and read back after the wipe process to assure that wiping has really taken place. This is especially important when firmware-based erase commands are used.
Attention is now directed to several drawings that illustrate features of the present invention.
Several drawings and illustrations are presented to aid in understanding the present invention. The scope of the present invention is not limited to what is shown in the figures.
The present invention is a system and method for first resetting (changing) the encryption key on a self-encrypting disk drive followed by a complete disk wipe.
The encryption key used in a self-encrypting hard disk drive is usually a long key used with a high-security encryption method like AES. This key is typically called the Media Encryption Key (MEK). This is a strong key generated automatically as a random or pseudo-random number by the disk hardware/firmware that is typically 128 or 256 bits. Some disk drives may use more than one MEK for different tracks or sectors.
Unlocking the drive for use may require another key typically called a Key Encryption Key (KEK) supplied by the user, BIOS, an operating system or a network. The MEK is encrypted by the KEK, and only the encrypted version of the MEK is stored when the drive is powered off. Also, in most systems, the KEK is never stored in plain text inside the drive. Some drives allow a mode where there is no KEK, or the KEK is not set. In this mode, the drive is always unlocked and appears not to be encrypting even though it is (using the MEK). If a KEK is set, the drive powers up locked (with the MEK only in encrypted form) until the correct KEK is given to the drive by the user.
When a locked self-encrypting drive is powered up, the BIOS typically first sees a shadow disk that is much smaller than the real disk. The shadow disk is usually around 100 megabytes and contains executable software. The software in the shadow disk is read-only and typically requests the KEK from the user to unlock the real disk for use and to decrypt the MEK so the real disk can be read and written to.
Usually, the shadow disk software stores a hash of the KEK so it can recognize if the user provides the correct KEK. When the user enters the correct pass code (either the KEK itself, or a password or other authentication) the shadow disk hashes that pass code or KEK and compares the hash with the stored hash of the KEK. If the two match, the MEK is decrypted using the KEK in what can be a symmetric or asymmetric encryption method, and puts the decrypted MEK into the symmetric encryption-decryption circuit inside the drive (without ever writing it to the magnetic or semiconductor medium). Usually, the BIOS is called from the disk to start again, but it now has the much larger real disk with a capacity in gigabytes rather than megabytes, and the operating system boots normally.
Every hard disk drive (magnetic or semiconductor) has an electrical interface to the computer or controller it is connected to. Most computers connect hard drives through various I/O channels. Every hard disk drive also has a set of commands that are generally executed by loading registers in the disk drive controller. In order to access the disk drive in order to sanitize it, the wipe hardware interface must electrically connect to the drive and be able to issue commands to the drive.
The interface 5 also controls authentication for issuing special commands such as a reset-key (cryptographic erase) command. Since, execution of this command generally renders all the data on the disk permanently unreadable, most systems require special, higher authentication in order to execute this command and other similar commands as opposed to simple read or write commands. In some systems, this command cannot be issued over the regular electrical interface. However, in most systems, commands of this sort can be issued by a higher authority than the user (in some systems called a crypto officer or the like). This is usually simply a user with a different password or a different KEK that must be entered. Authenticating under a lower authority user password only allows disk reads and writes and operational commands, while authenticating under the higher level password allows any operation including a key reset command. With almost all systems, there is no level of authority that can read out the plain text MEK or even the encrypted MEK.
Upon receipt of a reset-key command with the proper authentication, the interface 5 executes a special algorithm that generates a new, strong MEK of the required 128 or 256 bits. This is typically done with a pseudo-random number generator or the like. This new key is first encrypted with the KEK using the secondary encryption technique (which may be identical to the first), and the encrypted version of the MEK is stored on the shadow disk. The generated plain text MEK is than placed in the MEK hardware register 4. At this point, both the old MEK and its encrypted copy are permanently gone on most systems. The disk is still functional for reading or writing; however, any old data will not be readable. Any new written data is encrypted with the new MEK and can be read back with it. The process is almost transparent with the exception that all the old data is now just random bits.
A wipe operation can now begin. However, with self-encrypting disks, there is no way to force the medium write to a particular wipe pattern since all writes are encrypted by the MEK, and all MEKs are internally generated, strong keys. Thus, the actual patterns being written into the medium will be different from any pre-specified patterns. Also, each successive write of the same pattern (say 0×55 at a byte level) will become a different value as the encryption algorithm proceeds. Thus, each sector written with the same pattern will be totally different from every other sector written with that pattern.
Once authenticated, the wipe controller 9 sends either the reset key command to the drive, or begins to wipe it as the user wishes. If the particular wipe standard requires read back to verify that the original data has been wiped, that can also be performed. In this mode, a sector or other address is typically written followed by a read back. Some standards do not require read back in order to run faster. Also, some standards require that the entire wipe process be performed more than once (in some cases, up to three times). This can also be done.
The wipe controller 9 can also verify that a key reset has indeed taken place before beginning the wipe operation. This can be easily done by writing a known pattern to a predetermined sector (using the old MEK); issuing a key reset; and then reading back that sector (at that point under the new MEK). The result should be a collection of almost random bits and not the data that was written. This test also verifies that the encryption hardware is functioning, and that data is indeed being encrypted before being written.
The remote terminal or user interface 6 (which may be a cellular telephone) typically runs a graphical user interface (GUI) with menus and command selections known in the art. The remote terminal generally includes key data entry, a display screen which may be a touch screen and possible audio such as voice recognition and a speaker or earphones.
One possible problem with wiping that can occur is that all system indications show wiping has taken place, but in reality, it has not. This is especially true if the wiping of a storage device is based entirely upon a firmware-based erase or wipe command. It is entirely possible to issue such a command, receive an indication that the command has completed successfully, and then discover that either no wipe has actually occurred, or that only a partial wipe was performed (for example, the command aborted before completing without issuing any error indication).
For encrypted systems, it has previously been discussed how a key reset can be verified by writing known pattern to a predetermined sectors using the old key; issuing a key reset; and then reading back those sectors under the new key. While this has the effect of wiping, no real wiping has been done. Merely, the old encryption key has been destroyed. The original data encrypted with the old key is still in place. If an adversary had the old key, this data could still be decoded. This is generally why this operation is usually proceeded or followed by actual wiping.
For non-encrypted systems, verifiable known data patterns (which can be random or deterministic) can be written to strategic locations such as the first storage location, at periodic addresses within the address space, or at other known locations that do not require writing every address location. Particularly important locations are the first and last addresses in the device. In extremely high security wipes, verifiable known data can be written to every storage location on the device before any other operation takes place. While this acts as a pre-wipe, its purpose is to later verify that wiping has really taken place. Then, if a firmware-based erase command is used, or if a standard wipe process is used, it is possible to verify that wiping or erasure has indeed occurred.
For non-encrypted storage devices such as memories in most cellular telephones, flash drives and other storage device types, after the verifiable data is written in strategic locations, normal wiping takes place as previously described, or a firmware-based erase command can be issued. After wiping or erasure has completed, each of the strategic locations can be read back to verify that the known data is no longer present.
The same principle can be applied to encrypted devices, even if the actual encryption keys are not accessible. Verifiable data patterns can be written to the strategic locations under the old key (no key reset or key change is commanded). Then an entire disk wipe or firmware-based erase is applied to the device under the old key. The strategic locations are then read back, still under the old key, to make sure that the known data has changed. Finally, a key reset can be commanded. For most situations, this is sufficient; however, a second wipe or firmware-based erase under the new key can be performed for extra security, and even an additional key reset can be commanded after that. The written data after the first wipe is encrypted wipe patterns under the old key that may be random or repetitious. After the key change, all that can be read back is random data resembling noise. A second wipe under the new key leads to random or repetitious patterns that can be read back under the new key, but that have no meaning. A second key change renders even this data random. In any case, all of the original stored data is verified as being gone.
As previously stated, some devices are sufficiently erased by firmware-based erase commands. The method of the present invention provides a way to make sure such a command has indeed performed its job.
The present invention provides a way to conveniently secure and wipe multiple disks using a local or remote interface. In particular, the system can be controlled from a remote location over a network. The symmetric key (MEK) on a self-encrypting disk, magnetic or semiconductor storage device can be optionally reset before performing wipe operations. It can optionally be reset a second time after wipe operations for additional security. This renders even the wipe data inaccessible. The final result is one or more disks that can be optionally formatted and are ready and safe for use. The present invention also provides a way to verify firmware-based erase commands.
Several descriptions and illustrations have been provided to aid in understanding the present invention. One with skill in the art will realize that numerous changes and variations can be made without departing from the spirit of the invention. Each of these changes and variations is within the scope of the present invention.