Skip to content

Releases

v4.0.0-alpha.2

v4.0.0-alpha.2 Pre-release
Pre-release

Choose a tag to compare

@mr-tz mr-tz released this 21 Oct 11:01
a1f69f4

The second alpha pre-release of FLOSS 4.0: faster static analysis of large PE files, and more accurate ELF layouts.

Install with pip install --pre flare-floss (or pip install flare-floss==4.0.0a2), or download the standalone binaries below.

Changes since v4.0.0-alpha.1

  • Faster code analysis: python-lancelot and python-flirt 0.10.1 skip FLIRT matching when there are no signatures. Static analysis of large PEs is 1.7x to 5x faster. (#1511, williballenthin/lancelot#264)
  • lancelot 0.10.1 also fixes CFG, no-return and PE section-mapping bugs, and a panic on addresses outside all sections (#1345). #code tags can change on some PE files.
  • ELF: sections that pyelftools rejects (e.g. a .symtab whose size isn't a multiple of its entry size) are kept with their raw header instead of being dropped from the layout. (#1508)
  • ELF: empty relocation sections no longer hide #reloc tags. (#1508)
  • CLI: floss without arguments shows a short help message. (#1450)
  • Development: the pytest pre-commit hook only runs manually (#1507). Release artifacts include the license (#1506). Viewer dependency update (#1509).

JSON output can differ from alpha.1 in #code tags on some PEs and in the layout of ELFs with malformed section headers. The schema is unchanged.

Full Changelog: v4.0.0-alpha.1...v4.0.0-alpha.2

v4.0.0-alpha.1

v4.0.0-alpha.1 Pre-release
Pre-release

Choose a tag to compare

@mr-tz mr-tz released this 19 Oct 19:28
49b0edd

FLOSS v4.0.0-alpha.1

This is the first alpha pre-release of FLOSS 4.0, merging the QUANTUMSTRAND architecture into floss and unifying layout-aware static string analysis, semantic tagging, and deobfuscation into a single pipeline.

Install from PyPI with pip install --pre flare-floss (or pip install flare-floss==4.0.0a1), or download the standalone binaries below.

Highlights Since v3.1.1 / v3.2.0

Unified Layout-Aware Static Strings & Tagging

  • Layout analysis by default: FLOSS parses PE, ELF, and Mach-O (Thin and Universal/Fat) structure layouts—sections, segments, import/export/delay-import tables, resources, rich headers, relocations, code signatures, entitlements, and code ranges—and renders static strings within their structural context. Non-structured or empty inputs fall back to a single binary layout node.
  • Semantic string tags: Static strings are annotated using embedded databases and heuristics:
    • #capa — expert rules imported from capa
    • #winapi — Windows API and DLL names
    • #msvc — Microsoft C runtime strings (filtered by library false-positive threshold)
    • #common, #code-junk — global prevalence and instruction-byte junk (gp family; #code-junk is hidden by default)
    • #<library> — open-source library strings (oss family: #openssl, #zlib, #curl, etc.)
    • #code, #reloc, #duplicate, #decoded — layout and cross-category context tags
  • Filtering & views:
    • Filter by section (--section / --no-section), structure (--structure / --no-structure), tag or tag family (--tag / --no-tag), regex (--query), or per-section relevance cap (--max-strings).
    • --interesting hides noisy strings (#common, #duplicate, #code, #reloc, #code-junk) unless they also carry a highlight tag (#capa).
    • --summary emits a fast static-only overview (metadata, counts, tag histogram, and high-value strings).
    • --plain renders the classic flat string list without layout or tags.
    • --columns {tags,offset,structure,encoding} customizes the layout view columns.

Interactive Web Viewer & Standalone HTML Reports

  • --html output: floss --html sample.exe > report.html generates a self-contained offline HTML report with embedded results.
  • Hosted viewer: Upload floss -j JSON output to https://mandiant.github.io/flare-floss/ to interactively search, filter by tags/structures/types, inspect obfuscated and language strings, and download a standalone offline copy of the viewer.

Caching, CLI, and Security Improvements

  • Result caching: Analysis results are cached by default in the platform cache directory (configurable via FLOSS_CACHE_DIR, FLOSS_CACHE_ENABLE=0, FLOSS_CACHE_REFRESH=1).
  • Shell completions: Generate tab completions with floss --print-completion {bash,zsh,tcsh,fish,powershell}.
  • Safe Vivisect workspace loading: Crafted .viv files are no longer auto-loaded. Loading an existing <sample>.viv requires FLOSS_LOAD_VIV_WORKSPACE=1 (and saving requires FLOSS_SAVE_WORKSPACE=1).

Changes Since quantumstrand-beta3

  • Merged the separate qs preview tool into the main floss CLI and pipeline (floss.layout, floss.tags, floss.pipeline). The standalone qs entry point has been removed.
  • Unified the JSON results document (ResultDocument) to carry metadata, analysis, strings, and layout in a single schema.
  • Applied semantic tagging to classic strings (stack, tight, decoded, and language strings) in addition to layout strings.
  • Added --html standalone report generation, --summary, --plain, --interesting, --query, --max-strings, and --print-completion.
  • Replaced machofile with native Mach-O header parsing and switched expert regex matching to google-re2.
  • Hardened PE/ELF/Mach-O layout parsing (delay-load import table tagging, import name bounds checks, bounded resource recursion, raw PointerToRawData section placement, RVA clamping, security directory guards, large virtual extent caps, and empty/string-less input handling).
  • Updated global prevalence (gp) and open-source (oss) string databases, hid #code-junk strings by default, and applied the library false-positive threshold to #msvc.

Breaking Changes

  • Python >= 3.12 required: Dropped support for Python 3.8–3.11.
  • CLI flags:
    • Replaced --only and --no with --string-type and --no-string-type (static, stack, tight, decoded, language, all).
    • Replaced -functions / --functions with --analyze-functions.
    • Removed -l / --load; passing a saved JSON results file as the positional sample argument is now detected and loaded automatically.
    • Removed -x / --disable-progress.
  • JSON schema (floss -j):
    • Top-level document now includes layout (ResultLayout | null), metadata.{md5,sha1,sha256}, metadata.runtime.{layout,tags}, and analysis.{enable_language_strings,enable_layout,enable_tags}.
    • strings.static_strings entries now include tags, section, and structure.
    • In JSON mode (-j), CLI/pipeline errors are emitted as structured JSON ({"error": ..., "code": ...}) on STDERR.
  • Standalone binaries:
    • Standalone executables for Linux, Windows, and macOS bundle the tag databases, FLIRT signatures, and HTML viewer template. Note that --version in the PyInstaller standalone binary reports the git describe tag string (v4.0.0-alpha.1-0-g<sha>).

Full Changelog: v3.2.0...v4.0.0-alpha.1

v3.2.0

Choose a tag to compare

@mr-tz mr-tz released this 16 Oct 10:51

FLOSS v3.2.0

This release fixes a security issue in how FLOSS loads vivisect workspaces, adds basic ELF support, fixes several bugs including a crash on Rust binaries, and requires Python 3.10 or newer.

Security

Crafted .viv files could execute code when FLOSS analyzed a sample: either the sample itself named *.viv, or a <sample>.viv next to it. FLOSS now always parses samples as PE or ELF and only loads an existing <sample>.viv when FLOSS_LOAD_VIV_WORKSPACE=1 is set. Only enable this for .viv files you trust. The IDA Pro plugin follows the same rule.

Affected: v3.1.1 and earlier, and the QUANTUMSTRAND pre-releases.

New features

  • basic ELF support (#1071)
  • Ghidra import script migrated to PyGhidra (#1190)

Bug fixes

  • fix a crash rendering stack strings from Rust binaries (#1259)
  • no traceback on Ctrl+C (#1244)
  • raise an error on invalid decoded string types (#1154)
  • also check the location type in import taint analysis (#1243)
  • honor skip_libs when scoring calls to a function (#1242)
  • fix string type hints and related bugs (#1103)

Performance

  • faster is_all_zeros check (#1294)
  • more efficient rendering (#1235)

Breaking changes

  • Python 3.10 or newer is required (#1070)
  • FLOSS_SAVE_WORKSPACE only accepts 1/true/yes/on

QUANTUMSTRAND beta 3

QUANTUMSTRAND beta 3 Pre-release
Pre-release

Choose a tag to compare

@mr-tz mr-tz released this 22 Apr 10:52

This release marks a significant milestone for the QuantumStrand preview, introducing native Mach-O support for Apple binaries, expanded PE structure analysis, and a more refined terminal interface. We have also introduced a standalone qs command to streamline workflows.

🍎 Native Mach-O Support

QuantumStrand now provides deep visibility into Mach-O binaries, bringing the same level of structural analysis to macOS and iOS samples as it does for Windows PE files:

  • Architectures: Supports both Thin and Fat (Universal) Mach-O binaries.
  • Layout Mapping: Automatically identifies and maps segments (__TEXT, __DATA, etc.) and sections.
  • Security Metadata: Parses Code Signatures and Entitlements (plist) to surface binary permissions and security identifiers.

🔍 Expanded PE Structure Analysis

The PE engine in QuantumStrand has been enhanced to extract even more contextual data:

  • Export Table Parsing: Automatically identifies exported symbols and forwarders.
  • Rich Header Detection: Locates and maps the PE Rich Header structure.
  • Human-Readable Resources: PE resources are now mapped to descriptive names (e.g., "Dialogs", "Manifest", "Version Info") instead of raw IDs.

🖥️ Refined Terminal Rendering

We have overhauled the QuantumStrand output to be cleaner and more information-dense:

  • Tag Grouping: Consecutive strings with identical tags now utilize visual grouping bars, reducing output noise by up to 60% in highly tagged binaries.
  • Modern Aesthetics: Switched to light border primitives (─, │, ┐, ┘) for a less heavy-handed visual style.
  • Formatting Fixes: Improved tab rendering (4-space expansion) and centered headers.

🛠️ Developer & Power User Updates

  • New qs CLI: You can now run QuantumStrand directly via the qs command (e.g., qs suspicious.exe).
  • Performance Optimizations: Refactored OffsetRanges for faster overlap lookups and optimized memory slicing.
  • Expanded Test Suite: Added new unit tests and integration tests for Mach-O parsing and boundary conditions.

Dependencies

  • Introduced machofile for Apple binary support.
  • Introduced msgspec for high-performance JSON serialization.

Full Changelog: quantumstrand-beta2...v0.2.0

QUANTUMSTRAND beta 2

QUANTUMSTRAND beta 2 Pre-release
Pre-release

Choose a tag to compare

@mr-tz mr-tz released this 12 Jan 11:04

This release brings performance improvements and various other enhancements.

Features

  • Performance: Quantumstrand is now significantly more performant, especially on larger files. We've optimized how code and relocation regions are identified and checked by introducing a more efficient data structure for managing file offsets.
  • Version Flag: The quantumstrand CLI now includes a --version flag to display the current version.

Bug Fixes & Improvements

  • Dependency Updates: All project dependencies have been updated to their latest versions to include the latest features and security fixes.
  • New Test Workflow: A new workflow has been added to specifically run tests for the quantumstrand feature.
  • Expanded Test Coverage: We've added many new tests for quantumstrand's offset and code range detection, as well as for the core string extraction routines.

QUANTUMSTRAND beta 1

QUANTUMSTRAND beta 1 Pre-release
Pre-release

Choose a tag to compare

@mr-tz mr-tz released this 27 Aug 18:46

Beta 1 Release of FLOSS QuantumStrand (qs)

We are excited to announce the beta release of QuantumStrand (qs), a new experimental tool within the FLOSS suite designed to revolutionize how analysts interact with strings in binary files. While traditional floss excels at deobfuscating and extracting all possible strings, qs focuses on providing deep context and structure to the strings it finds.

image

What is QuantumStrand?

QuantumStrand is a static string analysis tool that extracts, tags, and presents strings within the hierarchical context of a file's structure. It's built for analysts who need to quickly understand where strings are, what they might be, and how important they are, without getting lost in a sea of undifferentiated text.

It's particularly effective for triage, allowing you to rapidly identify interesting regions and ignore common noise.

Key Features

  • Hierarchical Layout Analysis: qs doesn't just give you a flat list of strings. It parses PE files into a tree of structures - sections, headers, resources, overlays, and even embedded files - and shows you exactly where each string resides.

  • Intelligent String Tagging: Strings are enriched with contextual tags from various databases, helping you to quickly filter signal from noise. Tags include:

    • #common: Identifies common, globally prevalent strings that can often be ignored.
    • #winapi: Highlights Windows API and DLL names.
    • #<library_name>: Recognizes strings from known open-source libraries (e.g., #openssl).
    • #expert: Marks strings based on expert-curated rules for malware analysis.
    • #decoded: Indicates the file was likely XOR-encoded and has been automatically decoded.
    • #code & #reloc: Flags strings that overlap with executable code or relocation table entries, which are often false positives.
  • Automatic XOR Detection: qs can automatically detect and decode PE files that have been obfuscated with a single-byte XOR key.

  • Rich, Structured Console UI: The default output is a color-coded view of the file's layout. It uses borders and indentation to visually group strings by their containing structure, making it easy to navigate and understand the file's composition.

  • Detailed JSON Output: For tool integration and deeper analysis, qs produces a JSON document containing the full file layout, all extracted strings, their tags, offsets, and structural information.

Usage

To get started with QuantumStrand, simply run it against a binary file:

$ quantumstrand.exe /path/to/your/sample.exe

Command-Line Options

  • -n, --minimum-length <len>: Set the minimum string length (default is 4).
  • -j, --json: Output the results in JSON format instead of the rich console view.
  • -l, --load: Load and display a previously saved JSON results file.
  • -d, --debug: Enable verbose debugging output.

Try It Out!

QuantumStrand is still in beta, and we welcome your feedback! Please give it a try on your favorite samples and let us know what you think. Your input will be invaluable as we continue to develop and refine this new capability.

v3.1.1

Choose a tag to compare

@mr-tz mr-tz released this 26 Sep 13:13
3cd3ee6

The v3.1.1 release relaxes restrictions on dependencies, which enables you to use FLOSS with Python 3.12 and newer.

Full Changelog: v3.1.0...v3.1.1

v3.1.0

Choose a tag to compare

@mr-tz mr-tz released this 15 Apr 10:52
db9af41

New Features

Other Updates

New Contributors

Full Changelog: v3.0.1...v3.1.0

v3.0.1

Choose a tag to compare

@mr-tz mr-tz released this 12 Dec 10:28
3782dc9

This release fixes the missing language module in the v3.0.0 PyPI build.

v3.0.0

Choose a tag to compare

@mr-tz mr-tz released this 12 Dec 09:48
494372c

New Features

  • identification of programs written in Go, Rust, and .NET
  • extraction of strings embedded in Go programs
  • extraction of strings embedded in Rust programs

Other Updates

  • updates to the IDA plugin
  • upgraded minimum required Python version to 3.8
  • various bug fixes
  • various code quality improvements

Google Summer of Code 2023

@Arker123 contributed the majority of features and improvements during the Google Summer of Code working closely with the Mandiant FLARE team. We'd like to thank him for the great collaboration and discussions before, during, and after the twelve week program.

Contributors

Thanks to all our contributors, including @symbolicvoid, @DiegoRomeo, @sara-rn and especially @Arker123

Full Changelog: v2.3.0...v3.0.0