Securely wipe disk/Tips and tricks

See Securely wipe disk for the main article.

This article describes alternative wiping methods to the specialized utilities that can speed up wiping.

Warning: Reserved blocks will not be wiped by creation of the files but they can be disabled with utility tune2fs.

Wipe a single file

Warning: Wiping of single files is less effective, if the partition has been defragmented, resized or moved, or the files have been duplicated on the same device before. It is much harder to recover if the whole encrypted data container part was wiped but encryption will lower disk performance much, if it is not hardware based encryption that has almost no impact on the performance. See also Securely wipe disk#Data remanence.

Note: To preserve file access and modification time you can use touch(1) command to change time and stat(1) command to read time information before accessing the file. See also comparison of file systems for support of metadata and timestamps by them.

Wiping of a single file consists of two basic and one advanced anti-forensic time consumed method that can be done only with specialized tools, the last one method will not be covered in this article.

Overwrite with random data before deletion or replace content with another one without changing file size.
Wipe file and meta-data stored by the filesystem with specialized tools
- Search the whole disk for the deleted left-over parts of the file and wipe them too without making any changes to other files and their traces.

Overwriting of the file without changing its size can be done with common Linux utilities:

Invoking shred -x file will replace content of file with pseudo-random data without changing the filesize (-x). Using the -u option will remove file after overwriting it.
With mkfs you can convert file into the filesystem that will alter everything in it, mount and fill in with any other content for a better overwriting.
The dd will create a file with preset size and content of your chose, if destination file name exist then it will become overwritten. With dd command you can replace the whole file or only a part in it with another content by combining skip and seek options. You need to know size of the file to avoid expand of the file, to do it can use du -b file_name | cut -f1 or stat -c "%s" file_name. It is mandatory to use iflag=fullblock option, see dd#Partial read.
Replace content in a file with a single symbol to avoid size changing you can do with perl utility.

To wipe meta-data you can fill in partition with files that makes file system replace old entries about files with new or use specialized utilities for that, see wipe free space section below.

It is up to you how to combine all of Linux file creation and conversion tools to prevent recovery of files and/or mislead recovery tools and them who uses it by rewriting with random or replace with predefined content.

Note:

Simple overwriting has a chance to leave dislocated parts of file if free space is available.
To rewrite content of specific files without changing their location on the disk you can first create a file or files to make partition full and then rewrite or replace content in files you want to hide with preferred utilities. It consumes much useful time if free space is very big and files are very few.

Examples:

Perl command that will replace everything in the file with .:

$ perl -p -i -e 's\[^*]\.\g' file_name

dd:

$ source_content | dd bs=size_in_bytes count=1 iflag=fullblock of=destination_file seek=0

Or by using stdout redirection that works a slightly faster for creation but you will not be able to use seek option for skipping some parts in the destination:

$ source_content | dd bs=size_in_bytes count=1 iflag=fullblock > destination_file

Tip: If source file is lower than destination then you will need to combine it with the while loop described in the section below.

Overwrite the target

Prevent wiping mounted partitions

Choosing the device to be wiped needs extra care; a simple typo can be enough to damage the system. To minimize these risks, you can use a simple script to wrap your favourite wipe tool. For example:

if [[ -e "$1" && -b "$1" ]];then 
NOT_safe="$(lsblk -o "NAME,MOUNTPOINT" ${1//[0-9]/} | grep -e / -e '\]')";
 if [[ -z "$NOT_safe" ]];then
# Here you can use any of your favourite wiping tools 
# to wipe destination passed on command line and stored in variable "$1"
# 
  else
  echo 'Not allowed to destroy if any of the partitions is mounted: '"$NOT_safe"
 fi
fi

dd - advanced example

An alternative is to randomize the drive/partition using a randomly-seeded AES cipher from OpenSSL. For example:

DEVICE="/dev/sdX"
PASS=$(tr -cd '[:alnum:]' < /dev/urandom | head -c128)
openssl enc -aes-256-ctr -pass pass:"$PASS" -nosalt </dev/zero | dd obs=64K ibs=4K of=$DEVICE oflag=direct status=progress

The command above creates a 128 byte encryption key seeded from /dev/urandom. AES-256 in CTR mode is used to encrypt /dev/zero's output with the urandom key. Utilizing the cipher instead of a pseudorandom source results in very high write speeds and the result is a device filled with AES ciphertext.

The block size is set to 64K above as it is usually faster than the default 512 bytes. Experiment with larger block sizes to find the optimal transfer rate for your hardware: [1] and the references therein.

See also Dm-crypt/Drive preparation#dm-crypt wipe on an empty disk or partition for a similar approach.

Using a template file

Instead of zeros you can use a bunch of files you want to be found or partition prints made by mkfs formatted files but you should mount and fill it up with content or with any other repeated output from utilities of your choice.

One way is to wipe until device ends, but this type of redirection is not recommended because you have to use stop keys to break the while loop when errors about device end will show up:

$ while [ 1 -lt 2 ];do cat file1-to-use.as-template file2-to-use.as-template /tmp/templatefiles/* ;done > /dev/sd"XY"

With dd you can safely wipe repetitively without out-of-space-errors, if size to be wiped is set up correctly with options. By using dd inside the while loop for stdout you will be able to chose which part of the file you want to restore by combining the skip and seek options with random or fixed values e.g. restore only partition start or end from a file, related are head(1) and tail(1) commands for output of the file parts to stdout.

while [ 2 -gt 1 ]; do 
 if [ -z "$(pidof dd)"  ];then 
  break ;
 fi;
cat file1-to-use.as-template file2-to-use.as-template /tmp/templatefiles/* ;
done | dd of=/dev/sd"XY" seek=start_sector bs=sector_size count=sectors_to_wipe

Tip: For repeatedly rewriting of the same area. You can also use variables from second loop around while ... done | dd ... to change destination of the source template files to use on each new loop.

Note: If you have a real big amount of predefined data to use for overwriting then you can use the ability of SquashFS to compress files and folders. Compressing them into a mounted ramdisk will minimize the amount of physical reads but will higher CPU usage. Or create the compressed archive, copy into the ramdisk and use a file extraction utility that can extract to stdout. Downsides are that it is harder to navigate between folders and files, if you do not want to use all and same content for each run. Upsides are that you can use even raw parts of the archive, if it was split into volumes upon creation.

Wipe free space

Warning: It is not appropriate to use before preparations for block device encryption.

You can wipe free space by several ways:

Redirect output into a file instead of partition or device.

Create multiple file copies by using e.g. cp command in loops with random file names or destination directories until no free space will be left.

Use an utility that creates encrypted files with random password and file names. Some of the file compression utilities have options for compression methods, file types and can even split file into volumes of the preset size upon creation. By using some options randomly into a loop you will be able to fill the whole free space up with encrypted data and overwrite previous data.

Use a specialized program for the free space wiping such as:

wipefreespace — Wipe Free Space securely erases the free space on file systems to prevent recovery of deleted sensitive data.

https://sourceforge.net/projects/wipefreespace/ || wipefreespace^AUR

zerofree — Scans for non-zero free blocks in a filesystem and fills them with zeroes

https://frippery.org/uml/ || zerofree^AUR

Using dd

One can create a file that fills the empty space using dd:

dd if=source of=junk
sync
rm junk

The source can be the /dev/urandom or /dev/zero stream. The file is removed after making sure data has synchronized on disk.

Using 7-Zip

Password="$(dd if=/dev/urandom bs=128 count=1)"
DestinationFile="$((${RANDOM/0/1}$(date "+%s")/${RANDOM/0/1}))"
7z a -t7z -mhe=on -p"${Password}" -mx=0 -v1m ${DestinationFile} source

See also 7z(1) for description of used options.

The source can be a predefined file with random data or a device, e.g. /dev/urandom or another block device or partition on it, e.g. /dev/sd"XY", with data you are not afraid to be found then even deleted files on it will be compressed to the destination.

Note:

It is not necessary to set level of compression, enough to store data for minimizing CPU usage and a faster fill free space up.
If you are using a single file as a source then you can put it into the RAM disk or the /tmp folder, because it uses tmpfs that allocates some amount of RAM because it will speed up reading.

Create multiple files with help of the timeout command

The timeout command with randomized waiting time used it in a loop will break the command that will leave a file with random size. This is a slow method but is one of the possible alternatives. You can also use an array with predefined file names before the random part of it.

AA=${RANDOM/0/1};
timeout $((AA/100)) cat /dev/urandom > filename${RANDOM}.tmp;