nexus/internal/leviathan/binsnoop

binsnoop

A pure-Go binary analysis library for detecting and extracting embedded files, firmware images, compressed data, filesystems, and more from binary blobs. Ported from the binwalk Rust codebase with like-for-like scanning logic.

Zero external dependencies — everything is self-contained, including the Aho-Corasick multi-pattern matcher, CRC32 validation, and binary structure parsing.

Features

• 111 built-in signatures covering archives, compression, filesystems, images, executables, firmware headers, cryptographic constructs, and more
• Aho-Corasick multi-pattern scanning with overlapping match support
• Signature validation — every magic byte match is verified by a dedicated parser that checks headers, checksums, and structural integrity
• Confidence scoring — results are ranked Low / Medium / High so callers can filter noise
• Extraction support — carve identified objects out of a binary, or invoke external tools (7z, unsquashfs, etc.) to fully extract them
• Filtering — include/exclude signatures by name, restrict short signatures to offset 0, or add custom signatures
• JSON-friendly — all result types carry json struct tags for easy serialisation

Installation

go get binsnoop

Or add to your go.mod:

require binsnoop v0.0.0

Quick Start

package main

import (
    "fmt"
    "binsnoop"
)

func main() {
    scanner := binsnoop.New()

    // Scan a file
    results, err := scanner.ScanFile("firmware.bin")
    if err != nil {
        panic(err)
    }

    for _, r := range results {
        fmt.Printf("[0x%08X] %s (confidence: %d, size: %d)\n",
            r.Offset, r.Description, r.Confidence, r.Size)
    }
}

API Reference

Creating a Scanner

// Default scanner with all 111 signatures
scanner := binsnoop.New()

// Scanner with options
scanner, err := binsnoop.NewWithOptions(binsnoop.Options{
    Include:    []string{"gzip", "squashfs", "uimage"},  // only these signatures
    Exclude:    []string{"copyright"},                    // remove these signatures
    FullSearch: true,                                     // match short sigs at any offset
    AdditionalSignatures: []binsnoop.Signature{...},      // add custom signatures
})

Scanning

// Scan raw bytes
results := scanner.ScanBytes(data)

// Scan a file on disk
results, err := scanner.ScanFile("/path/to/firmware.bin")

Both return []SignatureResult sorted by offset.

Analysis (Scan + Extraction)

// Analyse raw bytes (with optional extraction)
analysis := scanner.AnalyzeBytes(data, "firmware.bin", true)

// Analyse a file on disk
analysis, err := scanner.AnalyzeFile("/path/to/firmware.bin", true)

Returns *AnalysisResults containing the file map and extraction outcomes.

Carving

Extract a single identified object from raw data:

err := binsnoop.CarveResult(data, results[0], "/tmp/output")

This writes the bytes at the result's offset/size to a file in the output directory.

Listing Signatures

names := binsnoop.ListSignatures()  // []string of all registered signature names

Scanner Info

scanner.SignatureCount()  // number of registered signatures (111)
scanner.PatternCount()    // total magic byte patterns across all signatures

Types

SignatureResult

type SignatureResult struct {
    Offset             int    `json:"offset"`              // byte offset in the input data
    ID                 string `json:"id"`                  // unique identifier for this match
    Size               int    `json:"size"`                // size in bytes (0 = unknown)
    Name               string `json:"name"`                // signature name (e.g. "gzip")
    Confidence         uint8  `json:"confidence"`          // 0=Low, 128=Medium, 250=High
    Description        string `json:"description"`         // human-readable description
    AlwaysDisplay      bool   `json:"always_display"`      // show even if low confidence
    ExtractionDeclined bool   `json:"extraction_declined"` // parser advises against extraction
    PreferredExtractor *Extractor `json:"-"`               // extractor to use for this result
}

Options

type Options struct {
    Include              []string    // restrict to these signature names
    Exclude              []string    // remove these signature names
    FullSearch           bool        // short sigs match at any offset
    AdditionalSignatures []Signature // user-defined signatures
}

AnalysisResults

type AnalysisResults struct {
    FilePath    string                       `json:"file_path"`
    FileMap     []SignatureResult             `json:"file_map"`
    Extractions map[string]*ExtractionResult  `json:"extractions"`
}

Confidence Levels

Constant	Value	Meaning
ConfidenceLow	0	Magic matched but weak validation
ConfidenceMedium	128	Header parsed and mostly valid
ConfidenceHigh	250	Full structural validation passed

Custom Signatures

You can add your own signatures at scanner creation:

mySignature := binsnoop.Signature{
    Name:  "my_format",
    Magic: [][]byte{[]byte("MYMAGIC")},
    Parser: func(data []byte, offset int) (*binsnoop.SignatureResult, error) {
        // Validate the match and return a result
        if offset+16 > len(data) {
            return nil, fmt.Errorf("insufficient data")
        }
        return &binsnoop.SignatureResult{
            Offset:      offset,
            Size:        1024,
            Description: "My custom format, size: 1024 bytes",
            Confidence:  binsnoop.ConfidenceHigh,
        }, nil
    },
    Description: "My custom file format",
}

scanner, _ := binsnoop.NewWithOptions(binsnoop.Options{
    AdditionalSignatures: []binsnoop.Signature{mySignature},
})

Supported Signatures

Archives (12)

gzip, zip, tarball, 7zip, rar, cab, deb, cpio, arj, iso9660, dahua_zip

Compression (10)

lzma, bzip2, xz, zstd, lz4, lzop, lzfse, zlib, compress'd

Filesystems (13)

squashfs, cramfs, jffs2, yaffs, romfs, ext, fat, ntfs, apfs, btrfs, logfs, ubi, ubifs

Images & Media (7)

png, jpeg, bmp, gif, svg, riff, pdf

Executables (2)

elf, pe

Firmware (32)

uimage, uboot, trx, tplink, tplink_rtos, dlob, arcadyan, chk, seama, cfe, jboot_arm, jboot_stag, jboot_sch2, packimg, binhdr, autel, android_sparse, android_bootimg, dtb, dlink_tlv, dlke, shrs, dkbs, dms, encfw, encrpted_img, wince, mh01, csman, dxbc, matter_ota, rtk

Cryptographic (17)

aes_sbox, aes_forward_table, aes_reverse_table, aes_rcon, aes_acceleration_table, rsa, openssl, gpg_signed, pem_certificate, pem_public_key, pem_private_key, pkcs_der_hash, luks, crc32, sha256, md5

Miscellaneous (18)

copyright, wind_kernel, vxworks_symtab, ecos, linux_kernel, linux_boot_image, linux_arm_zimage, linux_arm64_boot_image, efigpt, mbr, uefi_pi_volume, uefi_capsule, pchrom, pcapng, pjl, srecord, dmg, qnx_ifs, dpapi, qcow

Extraction

Many signatures have associated extractors that invoke external command-line tools. When you call AnalyzeFile or AnalyzeBytes with extract: true, binsnoop will:

1. Carve the identified data to a temporary file
2. Run the configured external tool (e.g. 7z x, unsquashfs, jefferson)
3. Return the extraction results including output directory and success/failure status

Common external tools used:

Tool	Signatures
7z	zip, cpio, arj, iso9660, xz, lzma, squashfs, dmg, efigpt
gzip	gzip
bzip2	bzip2
zstd	zstd
lz4	lz4
unsquashfs	squashfs
jefferson	jffs2
unyaffs	yaffs
tsk_recover	ext, fat, ntfs
uefi-firmware-parser	uefi_pi_volume, uefi_capsule, pchrom
simg2img	android_sparse
dtc	dtb

How It Works

Scanning Algorithm

1. Short-signature pass — signatures marked Short (e.g. zlib, MBR) are checked only at offset 0 (unless FullSearch is enabled)
2. Aho-Corasick scan — all magic byte patterns are compiled into an Aho-Corasick automaton for efficient multi-pattern matching. When a high-confidence sized match is found, scanning restarts from the end of that match
3. Validation — every magic match is passed to its signature parser which validates headers, checksums, and structure
4. Post-processing — results are sorted by offset, overlapping/duplicate matches are resolved (higher confidence wins), and unknown sizes are estimated based on the next significant match or EOF

Architecture

binsnoop.go       Public API (New, ScanBytes, ScanFile, AnalyzeFile, CarveResult, etc.)
scanner.go        Core scan engine (Aho-Corasick integration, post-processing)
match.go          Self-contained Aho-Corasick trie with overlapping match support
parse.go          Generic binary structure parser (u8/u16/u24/u32/u64, big/little endian)
types.go          Core types (SignatureResult, Signature, Extractor, Options, etc.)
common.go         Utility functions (CRC32, endian helpers, string helpers)
extract.go        Carving and external tool execution
magic.go          Signature registry — wires all 111 signatures together
sig_archives.go   Archive signature parsers
sig_compression.go Compression signature parsers
sig_filesystems.go Filesystem signature parsers
sig_images.go     Image/media signature parsers
sig_executables.go Executable signature parsers
sig_firmware.go   Firmware header signature parsers
sig_crypto.go     Cryptographic signature parsers
sig_misc.go       Miscellaneous signature parsers

License

This project is a Go port of binwalk by ReFirmLabs. See the original project for licensing details.