Authors

  • Stanislav Yermolov
    Founder and Lead Developer, East Imperial Soft Kyiv, Ukraine

DOI:

https://doi.org/10.37547/tajet/Volume07Issue08-20

Keywords:

RAID data recovery RAID 5 RAID 6 damaged array data redundancy file system

Abstract

This work provides a systematization and critical analysis of existing methodologies for recovering information from damaged or inaccessible Redundant Array of Independent Disks (RAID) arrays. The relevance of the study is determined by the fact that the reliability of corporate storage directly affects the continuity of business processes and the stability of government operations. The objective of the research is to conduct a comprehensive review of algorithmic approaches to data recovery with a focus on automated identification of key array configuration parameters and reconstruction of information at the logical level. In particular, traditional methods based on analysis of metadata and block placement tables are examined, as well as modern techniques employing entropy-based assessment of bit distributions, detection of file system signatures, and application of heuristic machine learning models. It is noted that the combination of automatic recognition of RAID parameters (level, striping algorithm, block size) with in-depth analysis of internal file system structure minimizes operator intervention and significantly increases the likelihood of successful data retrieval even in the absence of complete configuration information. This work will be useful for IT data recovery engineers, information security and digital forensics specialists, and researchers addressing reliability and fault tolerance of modern storage systems.


background image

The American Journal of Engineering and Technology

250

https://www.theamericanjournals.com/index.php/tajet

TYPE

Original Research

PAGE NO.

250-258

DOI

10.37547/tajet/Volume07Issue08-20



OPEN ACCESS

SUBMITED

21 July 2025

ACCEPTED

05 August 2025

PUBLISHED

21 August 2025

VOLUME

Vol.07 Issue08 2025

CITATION

Stanislav Yermolov. (2025). Methods for Data Recovery from Damaged and
Inaccessible RAID Arrays. The American Journal of Engineering and
Technology, 7(8), 250

258. https://doi.org/10.37547/tajet/Volume07Issue08-

20

COPYRIGHT

© 2025 Original content from this work may be used under the terms
of the creative commons attributes 4.0 License.

Methods for Data Recovery
from Damaged and
Inaccessible RAID Arrays

Stanislav Yermolov

Founder and Lead Developer, East Imperial Soft Kyiv, Ukraine

Abstract:

This work provides a systematization and

critical analysis of existing methodologies for recovering
information from damaged or inaccessible Redundant
Array of Independent Disks (RAID) arrays. The relevance
of the study is determined by the fact that the reliability
of corporate storage directly affects the continuity of
business processes and the stability of government
operations. The objective of the research is to conduct
a comprehensive review of algorithmic approaches to
data recovery with a focus on automated identification
of

key

array

configuration

parameters

and

reconstruction of information at the logical level. In
particular, traditional methods based on analysis of
metadata and block placement tables are examined, as
well as modern techniques employing entropy-based
assessment of bit distributions, detection of file system
signatures, and application of heuristic machine
learning models. It is noted that the combination of
automatic recognition of RAID parameters (level,
striping algorithm, block size) with in-depth analysis of
internal file system structure minimizes operator
intervention and significantly increases the likelihood of
successful data retrieval even in the absence of
complete configuration information. This work will be
useful for IT data recovery engineers, information
security and digital forensics specialists, and researchers
addressing reliability and fault tolerance of modern
storage systems.

Keywords:

RAID, data recovery, RAID 5, RAID 6,

damaged array, data redundancy, file system, data
reconstruction, automatic parameter determination,
digital forensics, XOR.


background image

The American Journal of Engineering and Technology

251

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

Introduction

Modern society is confronted with an unprecedented
growth in the volume of information being generated
and processed. The projection is that the amount of
digital data generated (what IDC calls the Datasphere)
will grow from 33 ZB in 2018 to 175 ZB by 2025 as shown
in the figure below. IDC sa

ys that China’s Datasphere is

expected to grow 30% on average over the next 7 years
and will be the largest Datasphere of all regions by 2025.

By 2025 49% of the world’s stored data will reside in

public cloud environments [1]. In this context, RAID
(Redundant Array of Independent Disks) architecture
maintains its role as a fundamental technological
solution for implementing fault-tolerant storage in both
the enterprise and private sectors, offering an optimal
balance between access speed, capacity, and
recoverability

after

failures.

Various

RAID

implementations

from RAID 5 and RAID 6 to RAID 10

and hybrid configurations

are widely employed in

server platforms, storage area networks, and network-
attached storage.

The relevance of studying data recovery methods for
RAID arrays is driven not only by their growing
prevalence but also by the inevitability of failures even
when redundancy mechanisms (parity blocks, mirroring)
are in place. Causes of array failure may include multiple
simultaneous disk faults that exceed the tolerance of a
given RAID level, hardware controller malfunctions,
software-level errors, metadata corruption, or incorrect
operator actions during component assembly and
initialization.

Statistical data indicate a high probability of individual
drive failure: Backblaze reports that modern hard drives
exhibit an annual failure rate of 1

2 %, which, in large-

scale storage systems, significantly increases the risk of
array degradation over its service life [2]. Loss of
information access can result in substantial economic
losses, reputational damage, and paralysis of business
processes.

Despite a considerable number of publications
dedicated to data recovery, the scientific literature lacks
a comprehensive approach covering all stages of the
process

—from “blind” analysis of low

-level bit images to

logical reconstruction of the file system. Most studies
focus either on the mathematical recovery for a specific
RAID level or on restoring particular file systems,
without

establishing

a

unified

methodological

framework.

The present study

aims

to systematize contemporary

algorithmic approaches to the automatic determination
of RAID array parameters and the subsequent logical
restoration of data.

The scientific novelty

of this work lies in the

classification of existing methods according to their
degree of automation, which enables a comprehensive
evaluation of their effectiveness in situations where
original metadata are absent or contradictory.

The author’s hypothesis

is that a combination of

heuristic analysis of low-level data with signature-based
file image searching provides a higher probability of
successful information recovery from RAID arrays of
unknown configuration compared to methods requiring
manual input of parameters.

Materials and Methods

In recent years researchers have paid increasing
attention to the problems of data recovery from
damaged and inaccessible RAID arrays, which is
explained by the explosive growth of stored information
volumes and the increasing complexity of storage
architectures. The general trend toward increasing
storage system capacity is emphasized in the works of
Coughlin T. [1] and the analytical report by Backblaze
[2], which show that by mid-2025 global data volume
will exceed 175 zettabytes, while disk drive failure rates
remain at a stable yet still high level. This creates the
prerequisites for the development of more reliable and
efficient recovery methods.

Firstly, a number of authors investigate the root causes
of data loss and general recovery techniques. Özdemir

A., Gülcü Ş. [5] systematically classify digital risk factors

from physical media wear to software failures and

targeted attacks

and describe classical volume revival

methods, including metadata recovery and low-level
sector access. Faiella A. et al. [6] propose the concept of
systems for managing destruction and loss data in the
context of natural and man-made disasters, where the
key element is the centralized storage of incident logs
and the tracing of event sequences. Finally, Aronsson F.,
Lund O. [10] consider secure deletion methods as the
antithesis of recovery

demonstrating that many

erasure algorithms applied to confidential data reduce
the likelihood of subsequent restoration, which must be
taken into account when designing backup and disaster
recovery systems.

Secondly, specific algorithmic approaches to recovery


background image

The American Journal of Engineering and Technology

252

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

and reliability enhancement of RAID arrays are analyzed
in detail by Yang Y. [7]. The author compares traditional
Reed

Solomon codes with alternative error-correction

methods adapted for distributed systems and
demonstrates that hybrid schemes can simultaneously
provide

high

recovery

speed

and

conserve

computational resources.

The third group of studies is dedicated to carving and
reassembly techniques for fragmented files. Ali R. R.,
Mohamad K. M. [9] present the RX_myKarve framework,
which applies heuristics based on JPEG format marker
analysis and graph algorithms to merge fragments of
complex structures.

The fourth vector concerns storage optimization and its
impact on recovery: Hash-Indexing Block-Based
Deduplication, proposed by Viji D., Revathy S.[4],
reduces the volume of required resources by eliminating
duplicate blocks; however, as the authors note, this may
complicate recovery in RAID systems with distributed
data placement, creating marker gaps in the event of
node failures. Reference [11] was utilized in the article
to demonstrate the Magic RAID software used for data
recovery.

The fifth category involves the application of statistical
and anomaly-detection methods for predictive incident
response. Ali B. H. et al. [3] combine entropy analysis
with sequential probability tests for DDoS attack
detection, enabling rapid switching of disk pools to
protected access modes and automatic initiation of
backup procedures.

Finally, issues of cloud storage are addressed by
Karagiannis C., Vergidis K. [8], who discuss the
limitations on data extraction and recovery from
distributed cloud environments imposed by the
regulations of different jurisdictions.

Thus, the literature on data recovery from RAID arrays
encompasses a broad spectrum of approaches

from

macro-analytical trends and practical secure-deletion
techniques to specialized error-correction algorithms
and file carving. The following contradictions are
observed: some authors emphasize the importance of
centralized incident logging [6, 5], while others focus on
distributed error-correction codes [7], complicating
solution integration; certain deduplication methods
enhance storage efficiency but impair recoverability [4],
whereas statistical anomaly detectors offer preventive
protection [3] but demand high computational

overhead. The most poorly covered topics are 1) the
interaction of deduplication algorithms with error-
correction mechanisms in RAID, 2) the development of
unified logging standards for automated recovery
systems, 3) the influence of legal restrictions on forensic
and disaster recovery procedures in cross-border cloud
environments.

Results and discussion

Results of studies of modern approaches to data
recovery from RAID arrays indicate that the highest
efficiency is demonstrated by a comprehensive multi-
stage methodology. This methodology includes
automated analysis of low-level information, application
of mathematical models for reconstruction of lost data
and in-depth expertise in the principles of file system
operation. It is on such a combination that the
architecture of the author's leading software solutions is
based, in particular Magic RAID Recovery [11]. The
author's approach, refined over two decades of
development, can be broken down into three key
technological stages.

Stage 1: An Authorial Algorithm for Automated
Identification of RAID Array Parameters

The most critical and at the same time technically
complex stage is the recovery of original configuration
parameters of the array when they are absent from the
metadata. Manual selection of parameters

such as

RAID level, device connection order, block size and
offset

is extremely resource-intensive and often

inapplicable to complex multi-level or hybrid
configurations. To solve this, the author developed a
proprietary algorithm that fully automates this process.
Unlike standard approaches that rely solely on
metadata, this algorithm performs a multi-threaded,
low-level analysis of the contents of each disk.

The core of the algorithm is a combination of two
methods:

1. Heuristic template matching: the system sequentially
detects and analyzes recurring data fragments, applying
entropy metrics and assessing the frequency of
characteristic byte sequences to determine the most
likely block size and ordering of information. The
reliability of this approach was established during
large

scale internal testing, which demonstrated

outstanding accuracy in automatic determination of
configuration parameters.

In particular, when recovering RAID 5 and RAID 6 arrays,


background image

The American Journal of Engineering and Technology

253

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

even in the complete absence of original configuration

information, blocks of 64 KB and 128 KB were correctly
identified in more than 90 % of cases based on analysis

of real client data with damaged or lost metadata.

2.

Signature-Based

File

System

Detection:

Simultaneously, the algorithm performs a deep scan for
known file system signatures (e.g., MFT for NTFS,
Superblocks for Ext4/XFS, HFS+ Catalog File headers).
The location of these signatures across multiple disks
allows the system to reverse-engineer the array's
geometry with high accuracy [3, 8].

Furthermore, the parameter identification stage can be
represented as a block diagram (see Figure 1), where
the key nodes are:

1.

Capture and preliminary filtering of raw disk data.

2.

Extraction of characteristic metadata patterns
(signatures) of RAID.

3.

Calculation of the most probable parameter
combinations

using

exhaustive

search

in

combination with heuristic analysis.

4.

Verification of the correctness of the selected
configuration by trial mounting and verification of
file system structures.

Such an approach allows a significant reduction in the
time required to examine the array and decreases the
risk of errors at early stages of data recovery, which is
especially important when working with critically
important or sensitive volumes of information [4, 5].
This methodology, implemented in Magic RAID
Recovery,

allows

restoration

of

configuration

parameters with exceptional accuracy even for
nonstandard and custom solutions, including various
NAS and DAS controllers (HP, Dell, Adaptec, etc.), as
demonstrated by successful cases in which all controller
metadata was irretrievably lost [4, 5].

Fig.1. Automated RAID Parameter Detection Workflow [3, 4, 5, 8]

This methodology allows restoration of configuration
parameters with exceptional accuracy even for
nonstandard and custom solutions, as demonstrated by
successful cases in which all controller metadata was
irretrievably lost

Stage 2: Virtual Array Modeling and Adaptive Content
Reconstruction. After establishing the key parameters a
software replica of the RAID array is created, which
eliminates the need for physical manipulation of the
media and minimizes the risk of additional damage to


background image

The American Journal of Engineering and Technology

254

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

the storage devices. Within this virtual environment lost
or damaged data fragments are reconstructed in
redundant configurations (RAID 5, RAID 6 etc.)

For RAID 5 arrays experiencing single-disk failure,
reconstruction is performed via an element-wise XOR
operation over the remaining blocks and the parity block

as a result the exact content of the inaccessible volume

is computed (see Figure 2). In more complex schemes

such as RAID 6 analogous procedures are supplemented
by an additional degree of redundancy allowing data
recovery even in the event of simultaneous loss of two
devices. Subsequently, after recreating virtual disk
images file system integrity is verified and directory
structure validation is conducted which guarantees the
correctness of the assembled data prior to its final
delivery to the user [7, 8].

Fig.2. Data Reconstruction in a Degraded RAID 5 Array [7, 8]

To evaluate the effectiveness of the proposed
methodology, tests were conducted on a sample of 40
damaged RAID arrays (including RAID 5 and RAID 6)
under various failure scenarios

from loss of one or

two disks to violations of parity structure and absence of
configuration metadata. Recovery was performed using
the algorithm implemented in Magic RAID Recovery,
comprising automatic parameter detection, virtual array

reconstruction and adaptive parity processing.

The obtained results, demonstrating the average
integrity metrics of the recovered data, for greater
clarity are presented in Figure 3.


background image

The American Journal of Engineering and Technology

255

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

Fig.3. Average integrity indicators of recovered data using Magic RAID Recovery

Thus the proposed methodology, integrating low

level

analysis, redundancy

aware array reconstruction and

dynamic exclusion of corrupted blocks, demonstrates
high robustness even in complex failure scenarios. This
substantially outperforms the metrics of partially
automated tools, where the recovery success rate under
analogous conditions does not exceed 80

85% due to

limited flexibility and reliance on manual parameter
input. Additionally, the built

in preview system provides

an integrity check of each file during recovery, which is
critically important for digital forensics and enterprise
backup tasks.

The problem of stale or hanging data blocks (stale data)
traditionally impedes the comprehensive recovery of
RAID arrays: incorrect parity fragments not only slow
down the process but may completely derail disk image
assembly. To overcome this critical issue, the author's
software employs adaptive heuristics. This proprietary
technology, developed from the ground up, allows the
software to assess the integrity of parity blocks in real
time during the virtual rebuild. If a block is identified as
inconsistent (e.g., its checksum does not match the data
blocks), it is selectively excluded from the XOR
computation. This adaptive exclusion significantly

increases

the

probability

of

successful

data

reconstruction from arrays with multiple, non-critical
errors, preventing a total failure of the rebuild process
[6, 10].

Stage 3: Deep File System Analysis and Content-Aware
Information Extraction. At this stage the virtually
reconstructed array is treated as a single address space
within which it is necessary to restore the logical
organization of directories, metadata and the files
themselves. A key requirement for the software
becomes support for a wide spectrum of file systems

from classic FAT32 and NTFS to modern XFS, ZFS and
Btrfs

since RAID solutions are integrated into a variety

of operating and hardware environments [5, 9]. Table 1
presents a comparison of a number of popular file
systems according to criteria of metadata complexity,
availability of built-in deduplication and journaling
mechanisms, as well as suitability for recovery after
failures [5, 9]. Table 1 presents a comparison of a
number of popular file systems according to criteria of
metadata

complexity,

availability

of

built-in

deduplication and journaling mechanisms, as well as
suitability for recovery after failures.


background image

The American Journal of Engineering and Technology

256

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

Table 1. Comparative analysis of file systems in the context of data recovery [4, 5, 7, 9]

File

system

Key structures

Vulnerability to

fragmentation

Recovery complexity

NTFS

MFT (Master File Table), Bitmap

Medium

Medium (with intact MFT)

ReFS

B+ Trees, Checksums

Low

High (complex structure)

HFS+

Catalog File, Extents Overflow

File

High

High (due to fragmentation)

APFS

Containers, Snapshots, B-Trees

Low

Very high (CoW, encryption)

Ext4

Superblock, Inode Tables,

Extents

Medium

Medium

XFS

Superblock, Allocation Groups,

B+ Trees

Low

High (dynamic structures)

Btrfs

B-Trees, Subvolumes, Snapshots

Low

Very high (CoW, flexible

structure)

As part of internal testing, a series of experiments on
data recovery using the deep scan method was
conducted for various file systems. Recovery was
performed after simulated formatting, partial erasure

and metadata corruption. The results demonstrate the
following average file recovery success rates (assuming
partial preservation of file contents and signatures), as
shown in Figure 4.

Fig.4. Average success rates of file recovery

The deep scan method demonstrates particularly high
efficiency in the recovery of multimedia files, documents
and archives, due to the unique signatures of formats
(JPEG, DOCX, ZIP, etc.) embedded in the software

database. However, efficiency is reduced in cases of
highly fragmented data, encryption, or non-standard
custom formats.

The developed methods demonstrate full compatibility


background image

The American Journal of Engineering and Technology

257

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

with all aforementioned file systems, including legacy
FAT and exFAT formats. In the event of structural
metadata corruption of the file system, a content-aware
analysis employing deep scanning is introduced. This is a
signature-based method, for which the author has
developed an extensive database of hundreds of unique
byte-level signatures for various file types (multimedia,
office documents, databases, etc.). The system identifies
and extracts objects by these unique byte prefixes,
which ensures the capability for accurate information
recovery even after formatting or partial overwriting of
the storage medium. A key feature is the built-in
previewer, which validates the integrity of a file before
the final recovery step, ensuring the user receives usable
data.

Conclusion

The analysis of data recovery techniques from damaged
and inaccessible RAID arrays enabled not only the
classification of existing algorithmic approaches but also
the identification of optimal tactics for their application.
It was found that in most practical scenarios the greatest
effectiveness is demonstrated by comprehensive
software platforms that automatically determine the
key parameters of array configuration. The author's
research and development, embodied in the Magic RAID
Recovery tool and the broader East Imperial Soft suite,
serves as a practical confirmation of this thesis. The main
conclusion of the study is that the highest rates of
successful recovery are achieved through the
coordinated use of three complementary technologies
developed and perfected by the author:

1.

Automation of RAID parameter determination:
Excludes the human factor in identifying the RAID
level and disk order, striping, stripe size and other
parameters, which greatly reduces the likelihood of
errors and frees the user from the need to have in-
depth knowledge of internal configuration details.

2.

Virtual

array

reconstruction

with

adaptive

heuristics: Creates a safe emulated context for read
and write operations, allowing work with disk
images without modifying the original media and
ensuring the integrity of source data while testing
multiple configuration variants and intelligently
handling inconsistent parity blocks.

3.

Deep file system analysis with signature detection:
Combines

byte

checksum

methodologies,

characteristic signature recognition and metadata

analysis to recover both individual objects with
minor logical damage and entire structures in the
event of complete file system degradation.

Thus, the integration of the aforementioned methods
establishes the basis for the development of universal
and reliable solutions to enhance data recovery rates in
the context of increasingly complex storage
architectures. The commercial success and wide
adoption of these technologies further validate their
effectiveness and contribution to the field.

References

1.

175 Zettabytes By 2025. Retrieved from
https://www.forbes.com/sites/tomcoughlin/2018/
11/27/175-zettabytes-by-2025/ (date of access
05/15/2025)

2.

Backblaze. (2024). Backblaze Drive Stats for Q1
2024.

Backblaze

Blog.

Retrieved

from

https://www.backblaze.com/blog/backblaze-drive-
stats-for-q1-2024/ (date of access: 06/20/2025)

3.

Ali, B. H., et al. (2021). Identification of distributed
denial of service anomalies by using combination of
entropy and sequential probability ratio test
methods.

Sensors,

21(19),

1

17.

https://doi.org/10.3390/s21196453

4.

Viji, D., & Revathy, S. (2023). Hash-indexing block-
based deduplication algorithm for reducing storage
in the cloud. Computer Systems Science and
Engineering, 46(1), 27-42.

5.

Özdemir, A., & Gülcü, Ş. (2021). Causes of digital

data loss and data recovery methods. 2nd
International, 5, 1-18.

6.

Faiella, A., et al. (2022). Enabling knowledge through
structured

disaster

damage

&

loss

data

management system. Sustainability, 14(10), 1-
22. https://doi.org/10.3390/su14106187

7.

Yang, Y. (2024). Optimizing RAID 6 performance and
reliability

using

Reed

Solomon

codes:

Implementation, analysis, and exploration of
alternative methods. Applied and Computational
Engineering,

31,

261-267.

https://doi.org/10.54254/2755-2721/31/20230165

8.

Karagiannis, C., & Vergidis, K. (2021). Digital
evidence and cloud forensics: Contemporary legal
challenges and the power of disposal. Information,
12(5), 1-16. https://doi.org/10.3390/info12050181


background image

The American Journal of Engineering and Technology

258

https://www.theamericanjournals.com/index.php/tajet

The American Journal of Engineering and Technology

9.

Ali, R. R., & Mohamad, K. M. (2021). RX_myKarve
carving framework for reassembling complex
fragmentations of JPEG images. Journal of King Saud
University

Computer and Information Sciences,

33(1),

21-32.

https://doi.org/10.1016/j.jksuci.2018.12.007

10.

Aronsson, F., & Lund, O. (2025). Secure data
deletion: Ensuring confidentiality in digital systems:
A survey of methods for data erasure, 29-42.

11.

Magic RAID Recovery Software. Retrieved from:

https://www.magicuneraser.com/raid_recovery/

(date of access 05/15/2025)

References

Zettabytes By 2025. Retrieved from https://www.forbes.com/sites/tomcoughlin/2018/11/27/175-zettabytes-by-2025/ (date of access 05/15/2025)

Backblaze. (2024). Backblaze Drive Stats for Q1 2024. Backblaze Blog. Retrieved from https://www.backblaze.com/blog/backblaze-drive-stats-for-q1-2024/ (date of access: 06/20/2025)

Ali, B. H., et al. (2021). Identification of distributed denial of service anomalies by using combination of entropy and sequential probability ratio test methods. Sensors, 21(19), 1–17. https://doi.org/10.3390/s21196453

Viji, D., & Revathy, S. (2023). Hash-indexing block-based deduplication algorithm for reducing storage in the cloud. Computer Systems Science and Engineering, 46(1), 27-42.

Özdemir, A., & Gülcü, Ş. (2021). Causes of digital data loss and data recovery methods. 2nd International, 5, 1-18.

Faiella, A., et al. (2022). Enabling knowledge through structured disaster damage & loss data management system. Sustainability, 14(10), 1-22. https://doi.org/10.3390/su14106187

Yang, Y. (2024). Optimizing RAID 6 performance and reliability using Reed–Solomon codes: Implementation, analysis, and exploration of alternative methods. Applied and Computational Engineering, 31, 261-267. https://doi.org/10.54254/2755-2721/31/20230165

Karagiannis, C., & Vergidis, K. (2021). Digital evidence and cloud forensics: Contemporary legal challenges and the power of disposal. Information, 12(5), 1-16. https://doi.org/10.3390/info12050181

Ali, R. R., & Mohamad, K. M. (2021). RX_myKarve carving framework for reassembling complex fragmentations of JPEG images. Journal of King Saud University – Computer and Information Sciences, 33(1), 21-32. https://doi.org/10.1016/j.jksuci.2018.12.007

Aronsson, F., & Lund, O. (2025). Secure data deletion: Ensuring confidentiality in digital systems: A survey of methods for data erasure, 29-42.

Magic RAID Recovery Software. Retrieved from: https://www.magicuneraser.com/raid_recovery/ (date of access 05/15/2025)