Analyzing Mft For Deleted File Recovery

Analyzing MFT for Deleted File Recovery

Overview

The NTFS Master File Table ($MFT) is the central metadata repository for every file and directory on an NTFS volume. Each file is represented by at least one 1024-byte MFT record containing attributes such as $STANDARD_INFORMATION (timestamps, permissions), $FILE_NAME (name, parent directory, timestamps), and $DATA (file content or cluster run pointers). When a file is deleted, its MFT record is marked as inactive (InUse flag cleared) but the metadata remains until the entry is reallocated by a new file. This persistence makes MFT analysis a primary technique for recovering deleted file evidence, reconstructing file system timelines, and detecting anti-forensic activity such as timestomping.

When to Use

• When investigating security incidents that require analyzing mft for deleted file recovery
• When building detection rules or threat hunting queries for this domain
• When SOC analysts need structured procedures for this analysis type
• When validating security monitoring coverage for related attack techniques

Detection Gaps & Validation

• Most-missed artifacts: a single $MFT parse misses where evidence actually survives - check MFT record slack (remnants of a prior file in the 1024-byte entry), resident $DATA of small deleted files (fully recoverable from the record itself), non-base records reachable via $ATTRIBUTE_LIST, and $MFTMirr. Always also pull $LogFile and $UsnJrnl:$J, which record deletes the MFT alone no longer reflects.
• USN journal wrap: $UsnJrnl:$J is a sparse, size-capped circular log - older deletions are silently dropped when it wraps, so "not in the USN journal" is not "did not happen." Corroborate with $LogFile transactions (DeallocateFileRecordSegment) and Volume Shadow Copies of $MFT for the pre-deletion state.
• Timestomping detection ($SI vs $FN): $STANDARD_INFORMATION times are settable from user mode (SetFileTime), but $FILE_NAME times are kernel-only. An $SI earlier than the $FN, zeroed sub-second (.000000000) precision, or $SI inconsistent with the MFT entry-number ordering implies timestomping - confirm against $LogFile/USN and the record's sequence number.
• Validate recoverability before claiming it: a non-InUse record with intact cluster runs is recoverable only if those clusters were not reallocated - verify against $Bitmap and carve the runs with icat, rather than trusting the recorded FileSize as proof of content.
• Interpretation pitfalls (false positives): archived/copied files and installers legitimately carry old $SI dates; NTFS file tunneling re-applies prior timestamps on recreate; reformat/defrag reorders MFT entries. Confirm volume timezone and clock skew, and cross-check a second artifact (Prefetch, Recycle Bin $I, Event Logs) before attributing a deletion to a user.

Prerequisites

• Forensic disk image (E01, raw/dd, VMDK, or VHDX format)
• MFTECmd (Eric Zimmerman) or analyzeMFT (Python-based)
• FTK Imager, Arsenal Image Mounter, or similar for image mounting
• Timeline Explorer or Excel for CSV analysis
• Python 3.8+ for custom analysis scripts
• Understanding of NTFS file system internals

MFT Structure and Record Layout

MFT Record Header

Each MFT record begins with the signature "FILE" (0x46494C45) and contains:

Key MFT Attributes

Deleted File Recovery Techniques

Technique 1: MFT Record Analysis with MFTECmd

# Extract $MFT from forensic image using KAPE or FTK Imager
# Parse the $MFT with MFTECmd
MFTECmd.exe -f "C:\Evidence\$MFT" --csv C:\Output --csvf mft_full.csv

# Filter for deleted files (InUse = FALSE) in Timeline Explorer
# Look for entries where InUse column is False

Identifying Deleted Files in CSV Output:

• InUse = False indicates a deleted or reallocated record
• ParentPath shows original file location before deletion
• FileSize shows the original size (may still be recoverable)
• Timestamps in $STANDARD_INFORMATION and $FILE_NAME attributes persist

Technique 2: USN Journal ($UsnJrnl:$J) Analysis

The USN Journal records all changes to files on an NTFS volume, including creation, deletion, rename, and data modification events.

# Parse USN Journal with MFTECmd
MFTECmd.exe -f "C:\Evidence\$J" --csv C:\Output --csvf usn_journal.csv

# Key USN reason codes for deletion evidence:
# USN_REASON_FILE_DELETE     = 0x00000200
# USN_REASON_CLOSE           = 0x80000000
# USN_REASON_RENAME_OLD_NAME = 0x00001000
# USN_REASON_RENAME_NEW_NAME = 0x00002000

Technique 3: $LogFile Transaction Analysis

The $LogFile stores NTFS transaction records that can reveal file operations even after the USN Journal has been cycled.

# Parse $LogFile with LogFileParser
LogFileParser.exe -l "C:\Evidence\$LogFile" -o C:\Output

# Look for REDO and UNDO operations indicating file deletion:
# - DeallocateFileRecordSegment
# - DeleteAttribute
# - UpdateResidentValue (clearing InUse flag)

Technique 4: MFT Slack Space Analysis

MFT slack space exists between the end of the used portion of an MFT record and the end of the allocated 1024 bytes. This area may contain remnants of previous file records.

import struct

def parse_mft_slack(mft_path: str, output_path: str):
    """Extract and analyze MFT slack space for deleted file remnants."""
    with open(mft_path, "rb") as f:
        record_size = 1024
        record_num = 0
        slack_findings = []

        while True:
            record = f.read(record_size)
            if len(record) < record_size:
                break

            # Verify FILE signature
            if record[:4] != b"FILE":
                record_num += 1
                continue

            # Get used size from offset 0x18
            used_size = struct.unpack("<I", record[0x18:0x1C])[0]

            if used_size < record_size:
                slack = record[used_size:]
                # Check if slack contains readable strings or attribute headers
                if any(c > 0x20 and c < 0x7F for c in slack[:50]):
                    slack_findings.append({
                        "record": record_num,
                        "used_size": used_size,
                        "slack_size": record_size - used_size,
                        "slack_preview": slack[:100].hex()
                    })

            record_num += 1

    return slack_findings

Correlation with Supporting Artifacts

Cross-Reference MFT with $Recycle.Bin

# Parse Recycle Bin with RBCmd
RBCmd.exe -d "C:\Evidence\$Recycle.Bin" --csv C:\Output --csvf recycle_bin.csv

# Correlate: $I files contain original path and deletion timestamp
# Match MFT entry numbers from $R files back to original MFT records

Cross-Reference MFT with Volume Shadow Copies

# List volume shadow copies
vssadmin list shadows

# Mount shadow copies and extract $MFT from each
# Compare MFT records across shadow copies to track file changes over time

Forensic Value

• Deleted file metadata recovery: Original filename, path, size, and timestamps
• Timeline reconstruction: File creation, modification, access, and deletion events
• Timestomping detection: Comparing $SI vs $FN timestamps
• Data carving guidance: MFT cluster runs point to file content on disk
• Anti-forensic detection: Identifying wiped or manipulated MFT records

References

• NTFS MFT Advanced Forensic Analysis: https://www.deaddisk.com/posts/ntfs-mft-advanced-forensic-analysis-guide/
• MFT Slack Space Forensic Value: https://www.sygnia.co/blog/the-forensic-value-of-mft-slack-space/
• MFTECmd Documentation: https://ericzimmerman.github.io/
• SANS FOR500: Windows Forensic Analysis

Example Output

$ MFTECmd.exe -f "C:\Evidence\$MFT" --csv /analysis/mft_output

MFTECmd v1.2.2 - MFT Parser
==============================
Input: C:\Evidence\$MFT (Size: 384 MB)
Total MFT Entries: 395,264

Parsing MFT entries... Done (12.4 seconds)

--- Deleted File Recovery Summary ---
Total Entries:          395,264
Active Files:           245,832
Deleted Files:          149,432
  Recoverable:          87,234 (resident data or clusters not reallocated)
  Partially Recoverable: 31,456 (some clusters overwritten)
  Unrecoverable:        30,742 (all clusters reallocated)

--- Recently Deleted Files (Incident Window: 2024-01-15 to 2024-01-18) ---
MFT Entry | Filename                          | Path                               | Size      | Deleted (UTC)         | Recoverable
----------|-----------------------------------|------------------------------------|-----------|-----------------------|------------
148923    | exfil_tool.exe                    | C:\ProgramData\Updates\            | 1,258,496 | 2024-01-17 02:45:12   | YES
148924    | exfil_tool.log                    | C:\ProgramData\Updates\            | 45,312    | 2024-01-17 02:45:14   | YES
149001    | passwords.txt                     | C:\Users\jsmith\Desktop\           | 2,048     | 2024-01-17 02:50:33   | YES
149150    | scan_results.csv                  | C:\Users\jsmith\AppData\Local\Temp | 892,416   | 2024-01-17 03:00:01   | PARTIAL
149200    | mimikatz.exe                      | C:\Windows\Temp\                   | 1,250,816 | 2024-01-18 01:15:22   | YES
149201    | sekurlsa.log                      | C:\Windows\Temp\                   | 32,768    | 2024-01-18 01:15:25   | YES
149302    | .bash_history                     | C:\Users\jsmith\                   | 4,096     | 2024-01-18 03:00:00   | NO
149400    | ClearEventLogs.ps1                | C:\Windows\Temp\                   | 1,536     | 2024-01-18 03:01:12   | YES

--- $STANDARD_INFORMATION vs $FILE_NAME Timestamp Analysis (Timestomping Detection) ---
MFT Entry | Filename            | $SI Created          | $FN Created          | Delta     | Verdict
----------|---------------------|----------------------|----------------------|-----------|----------
148923    | exfil_tool.exe      | 2023-06-15 10:00:00  | 2024-01-15 14:34:02  | -214 days | TIMESTOMPED
149200    | mimikatz.exe        | 2022-01-01 00:00:00  | 2024-01-16 02:30:15  | -745 days | TIMESTOMPED

Recovered files exported to: /analysis/mft_output/recovered/
Full CSV report: /analysis/mft_output/mft_analysis.csv (395,264 rows)
Timeline CSV: /analysis/mft_output/mft_timeline.csv