- New
--format sarifoutput produces OASIS SARIF 2.1.0 documents compatible with GitHub Code Scanning, Azure DevOps, GitLab Security, and mainstream SIEM and SOAR pipelines - Anya's 15 rule catalogue (ANYA-V001 verdict carrier plus 14 signal rules) is embedded in every SARIF run, so consumers see human readable descriptions alongside each finding
- MITRE ATT&CK taxonomy emitted as
taxonomies[]when the analysis attaches any techniques, with each taxon linked to the official MITRE page - Per finding
properties.tags[]use a colon prefixed namespace vocabulary (verdict, mitre, family, confidence, signal, format) for SIEM filtering --output report.sarifwrites the SARIF document to a file instead of stdout- Docker images now published to GitHub Container Registry at
ghcr.io/elementmerc/anyain addition to Docker Hub - Reference
docker-compose.ymlgains ananya-watchservice demonstrating the sidecar inbox pattern (continuous directory watch with JSON verdict emission) - Releases now publish a consolidated
SHA256SUMS.txtfile covering every binary asset; verify withsha256sum -c SHA256SUMS.txt. Individual.sha256sidecars also still ship for backward compatibility. anya --file -reads bytes from stdin (size cap 1 GiB), enabling pipeline composition such ascat sample.exe | anya --file - --format sarif
- Bug fixes and improvements
- Fewer false positives on legitimate security tools and development utilities (FLOSS, tooling suites, dual use binaries)
- Better handling of JARs, Python wheels, Chrome extensions, APKs, and Office OOXML documents, which no longer flag as suspicious simply for containing executable entries
- More accurate compiler and toolchain identification for unsigned binaries, reducing noise on small MSVC utilities, Apple tooling, and MinGW compiled programs
- Reduced noise on ELF shared libraries and relocatable object files, which no longer trigger hardening detections that only apply to executables
- New detection signals for Batch and CMD scripts: LOLBin usage (certutil, bitsadmin, mshta, regsvr32, and similar), Windows Defender tampering patterns, hidden execution flags, and privilege escalation bypass patterns
- New entry point signature matching for packed PE binaries, with coverage for UPX, ASPack, Themida, MPRESS, PECompact, FSG, Enigma, and VMProtect
- Known sample lookups now short circuit to a clean verdict for curated tools and test files, so legitimate analysis utilities no longer surface as suspicious because of their metadata profile
--no-ksdand--ksd-thresholdflags now work (previously accepted but ignored); use--no-ksdto see the pure heuristic verdict and--ksd-threshold Nto tighten or loosen the similarity cutoff
- Bug fixes and improvements
- 7 new format parsers (VHD, OneNote, IMG, RAR, GZIP, 7-Zip, TAR)
- Secrets detection (AWS keys, JWTs, private keys, API tokens)
- Kernel driver analysis (.sys files)
- Improved scoring for HTA, JavaScript, VBScript, shell scripts, ISO, and XML
- Scoring engine recalibrated for improved detection across all file formats
- Known sample verdicts for security tools, development utilities, and test files
- Malware family context with descriptions and aliases for 50+ families
- Notes section on Overview tab for known sample and fragment annotations
- Teacher Mode guidance for Graph tab with interactive node explanations
- Threat path highlighting in batch graph (click to trace connected clusters)
- 260 DLL explanations and 8 API category explanations in Teacher Mode
- Proprietary explanation data served via IPC
- JSONL streaming output (
--jsonl) for pipeline integration - YARA only scanning mode (
--yara-only) for fast rule testing - Configurable analysis depth (
--depth quick|standard|deep) - Exit code from verdict (
--exit-code-from-verdict) for CI/CD integration - Watch mode JSON output (
anya watch --json) for structured monitoring - Coloured verdict output for TOOL, PUP, and TEST classifications
- Bug fixes and improvements
- Calibrated scoring engine: 99.9% detection, 0.1% FP on 10,320 samples
- Comprehensive known-product suppression (MinGW, .NET, Wine, Node.js, Ghidra tools)
- Python exec+subprocess scoring refined (exec+network = High, exec+subprocess = Medium)
- Archive format suppression for ZIP/JAR/7z entropy and IOC volume
- DLL weighted verdict threshold raised to reduce structural heuristic FPs
- Dedicated Graph tab with Obsidian-style 2D force-directed visualization
- Single-file IOC/Import evidence web (DLLs, APIs, IOCs, categories as nodes)
- Batch TLSH relationship graph with cluster halos and search-to-highlight
- Constellation emerge animation, smooth hover spotlight with alpha interpolation
- Edge labels on hover, drag-and-release (nodes spring back)
- Tab state persistence (scroll position and search filters preserved)
- Animated SVG empty state icons matching tab lucide icons
- Folder drag-and-drop triggers batch analysis automatically
- Batch sidebar search bar for filtering file list
- Teacher Mode for all Security tab cards and Graph tab
- Keyboard shortcut hints in New Analysis menu
--json-compactflag for single-line JSON output
- Bug fixes and improvements
- 20 format parsers (PE, ELF, Mach-O, JS, PowerShell, VBS, Batch, Python, OLE, RTF, ZIP, HTML, XML, Image, LNK, ISO, CAB, MSI, PDF, Office)
- YARA-X integration with rules loaded from
~/.config/anya/rules/ - Compiler/toolchain fingerprinting and import behavioural clustering
- Known-product suppression, benign IOC marking, byte histogram analysis
- TLSH Known Sample Database with forensic fragment annotation
- 99.6% detection, 0.8% FP on ~7,100 samples
- Force-directed network graph (batch + single-file modes)
- Format Analysis tab with 17 format-specific cards
- YARA matches, forensic fragments, toolchain detection in Security/Overview tabs
- PDF export, histogram flatness, benign IOC marking
anya benchmark— detection rate and performance benchmarking--format pdfand--format markdownreport output- Batch
--summaryshows TLSH similarity relationships
- PDF, Markdown, and HTML report generation
- All formats include KSD match, MITRE ATT&CK, analyst findings
- Bug fixes and improvements
- Improved detection accuracy across PE, ELF, and Mach-O formats
- Fallback PE import parser for packed/malformed import tables
- IOC volume now contributes to verdict
- ELF signals now feed into verdict (GOT/PLT, interpreter, RPATH, W+X, security features)
- Mach-O signal amplification for unsigned binaries
- Weighted verdict thresholds for signal accumulation
- Unrecognized file formats now check signals before returning UNKNOWN
- Refactored scoring engine into modular architecture
- Consolidated detection patterns into dedicated scoring module
- Eliminated duplicate scoring implementation in Tauri IPC layer
- Dockerfile: fixed missing anya-stubs/proprietary COPY, sed pattern, env var
- Release workflow: trimmed from 17 to 11 assets (removed linux-gnu, arch-specific macOS, NSIS exe, app.tar.gz)
- CI: added Docker build validation and musl cross-compile jobs
- CI: runs on dev pushes only (not main) to reduce redundant builds
- Tauri targets: appimage, deb, rpm, dmg, msi (removed NSIS and app.tar.gz)
- Fixed piped
curl | bashsilently defaulting to CLI-only (now prompts via /dev/tty) - Added --cli, --gui, --both flags
- Changed default to "both" (was "cli")
- Added --help, platform context in prompt, post-install launch hints
- Guided tour: Skip button now readable, tooltips clamped to viewport
- Chocolatey community package scaffolded (.choco/ directory)
- Internal testing improvements
- Case management — save, browse, and organise investigation cases (CLI + GUI)
- Watch mode — auto-analyse new files dropped into a directory
- Compare mode — side-by-side diff of two file analyses
- HTML report export
- Batch analysis in GUI with summary dashboard
- Keyboard shortcuts and drag-and-drop tab reorder
- IOC detection (IPv4, IPv6, URL, domain, email, registry, paths, mutex, Base64)
- File type mismatch detection
- Confidence scoring on all detections
- 15 new Teacher Mode lessons
- Private repo split for scoring engine and educational data
- Bug fixes and improvements
- Custom uninstaller wizard
- DLL and function explanations in Imports tab
- Draggable Teacher Mode sidebar
- Native installer branding for all platforms
- RPM packaging support
- Bug fixes and improvements
- First-run installer wizard
- Splash screen with theme adaptation
- Sliding tab indicator and settings animations
- Docker fixes (binary path, runtime image, permissions)
- Bug fixes and improvements
- PE and ELF static analysis with MITRE ATT&CK mapping
- Packer, compiler, and anti-analysis detection
- Entropy analysis with per-section breakdown
- Desktop GUI (Tauri v2 + React) with 7 analysis tabs
- Teacher Mode — contextual lessons for learning while analysing
- CLI with batch scanning, JSON output, and configuration
- Docker support, Debian packaging, one-liner install script
- Bible verse status bar (CLI + GUI)
- ELF binary analysis
- PE extended analysis (imphash, Rich header, TLS callbacks, overlay)
- MITRE ATT&CK tab in GUI
- Debian/Kali packaging and install script
- Bug fixes and improvements
- Testing infrastructure and CI/CD pipeline
- Library API, examples, and benchmarks
- Bug fixes and improvements
- Batch directory scanning
- Configuration file support
- JSON and file output modes
- Progress bars
- PE parsing, import/export analysis, suspicious API detection
- Security feature detection (ASLR, DEP/NX)
- Per-section entropy calculation
- Initial release — hashes, strings, entropy for PE files