Entity
Zero Breach Toolkit
Last updated 1 November 2026
-
Overview
Zero Breach Toolkit is advertised as a free, open-source checker for published traces of mercenary spyware, including Pegasus, Predator, Graphite and Candiru, on computers, Android evidence and iPhone backups.[1] Its official GitHub project is
interplains/Zero-Breach-Toolkit, and its README credits archivegenocide and identifies the separate Zero Breach project as its foundation.[1][2] It is a triage tool rather than a full forensic examination: unpublished spyware, expired traces and missing evidence can escape its checks.[1] -
Key facts
- Official project and identity: The repository is
interplains/Zero-Breach-Toolkit; the GitHub profile displays “ArchiveGenocide” and the username “interplains.” The project says its maintainers are anonymous and does not have named, publicly identified maintainers and approvers.[1][3][4] - Purpose: The toolkit compares supported local artifacts with known spyware indicators; it does not establish that a device is free of spyware or replace expert forensics.[1][1]
- Audit status: The available third-party evidence consists of limited code, documentation, network and runtime-preview reviews—not a comprehensive independent security audit or measured detection rate.[5][6][7][7][5]
- Reviews: Agentbeam inspected v0.1.7 code and documentation, Elite Performance Exclusive checked a v0.1.10 checksum and network behavior, and DEV author prodbymarcu examined v0.1.8 runtime modules and a preview without completing a full host scan.[5][6][7][7]
- Community coverage: GeekNews published a toolkit summary, whereas the located Hacker News submission links the narrower
interplains/zero-breachproject; neither is evidence of a toolkit security audit.[8][9][10] - Privacy boundary: A scan makes no intentional network requests and has no upload feature, but its Python socket and local-path guards are not an operating-system network sandbox. Strict offline execution requires operating-system network isolation.[11]
- Read-only boundary: Scanning does not modify source evidence or system settings, but it writes temporary database snapshots and requested reports; it does not quarantine evidence or install software.[2][11]
- Release verification: Signatures, checksums and build attestations address provenance and integrity, not freedom from bugs or proof of safety. Native desktop programs are not Microsoft- or Apple-signed.[2][4]
- Important phone-result correction: The v2.0.4 release advises users of v2.0.0–v2.0.2 to update and recheck phones previously shown as green because those versions could overstate what had been checked.[12]
- No research-organization endorsement: The project explicitly disclaims affiliation, endorsement or review by the research organizations whose published work it uses.[13]
- Official project and identity: The repository is
-
Evidence
Project, distribution and licensing
Item Supported details Main repository interplains/Zero-Breach-Toolkit; the runnable entry point isscan/zero-breach-toolkit.py, and scanner implementation is underscan/zbtriage/, not asrcfolder. The package retains its earlier name for existing installations.[1]Attribution The README says “Made by archivegenocide” and “Built on Zero Breach.” The NOTICE identifies copyright 2026 archivegenocide.com and says both projects share authors and licence.[2][13] Related Zero Breach project interplains/zero-breachis a separate, narrower local computer scanner for indicators of the AdFection/Candiru ad-delivered spyware campaign.[9]Companion website The website says it is based solely on the project README, directs software, documentation, updates and issues to the project and its maintainers, and disclaims responsibility for the project and software use.[14] SourceForge Its project page identifies itself as an exact mirror and explicitly states that SourceForge is not affiliated with Zero Breach Toolkit.[15] Windows desktop asset zero-breach-toolkit-windows-x64.exe; the README describes a double-click launch and warns that Windows may display a protection warning.[2]macOS desktop assets Separate Apple Silicon and Intel DMGs; the README requires launching the app on the Mac desktop rather than through a remote or SSH session.[2] Linux desktop asset zero-breach-toolkit-linux-x64.tar.gz, extracted and launched as./zero-breach-toolkitor through a terminal launcher.[2]Code licence Code, including the generated scanner, is Apache License 2.0.[1] Indicator licence Curated files in iocs/and generated indicator exports indist/are CC0 1.0.[1]Other licensing Optional research rules retain their own licences; lists downloaded through --updateretain their publishers’ terms and are not part of the repository.[1]Default Python dependencies The default scanner needs only Python’s standard library. Optional pattern scanning needs Python 3.9+ and the pinned official YARA-X 1.21.0 wheel.[16] Optional encrypted-backup dependencies Source-version iPhone backup decryption needs the owner’s password and the optional dependency in requirements-backups.txt; nothing is installed automatically.[17]Scan scopes and evidence coverage
Scope or evidence What is checked and important limits Quick Browsers only; the README estimates 5–10 minutes.[2] Standard Browsers, apps, startup items, program files and out-of-date programs; the README recommends this scope and estimates 15–20 minutes.[2] Full Adds web caches and more file types; the README estimates 30 minutes or more. Full mode does not cover every drive or account.[2][18] Published indicators New or unpublished spyware is not found; spyware may be rebuilt for individual targets, and many traces expire within weeks.[1] Indicator files Curated sets, downloaded feeds and STIX bundles are supported. Missing, refused, stale and undated sources are reported; a source update dropping its count by more than half or to zero is refused unless explicitly accepted with --update --accept-shrink SOURCE_ID.[18]Chromium browser families Chrome, Edge and Brave—including the documented stable, beta, dev, canary and nightly channels—plus Chromium, Vivaldi, Opera, Opera GX, Arc and Yandex.[18] Firefox browser families Firefox, LibreWolf, Waterfox, Floorp, Zen, Pale Moon and Tor Browser; portable installations elsewhere are not automatically discovered.[18] Safari Supported on macOS only.[18] Chromium and desktop-app artifacts History, bookmarks, cookies, favicons, site storage, connection records, HSTS and download URL chains; selected browser caches and raw app caches in Full mode. Locked files, deleted data and changed schemas limit coverage.[18] Browser-result interpretation Bookmarks and raw cache strings are references; download status and received bytes affect whether an entry is attempted or contacted, but neither establishes execution.[18] File fingerprints Platform-specific user, application and temporary roots; Windows candidate extensions; Linux/macOS executable or extensionless files, native libraries and scripts; all extensions in Full or targeted mode. Matching is limited to exact known hashes, with a 512 MiB per-file cap and inaccessible-path gaps.[18] Windows host artifacts DNS, COM, download origins, Win+R, Defender, PowerShell event logs, ShimCache and PCA. Logging can be disabled and records can expire; ShimCache indicates presence rather than execution.[18] Additional Windows records PSReadLine, Explorer TypedPaths, Run/RunOnce, services, Startup files with supported .lnkmetadata, task actions, Prefetch and Amcache. Sysmon and Security events are opt-in; BAM and SRUM are unsupported.[19][18]Windows permissions Ordinary-account checks are described for the default evidence readers; --extended-windowsSysmon and Security event readers may require administrator rights.[19]Windows antivirus status Windows Security Center product names, states and signature status, without a vendor allowlist. Windows Server and unregistered products are excluded; reported status is not an independent health test or proof of past protection.[18] Linux startup artifacts Standard user/system autostart .desktopfiles, systemd units, one-level.d/*.confoverrides, readable cron configuration, BusyBox/etc/crontabs, and hourly/daily/weekly/monthly script directories. Links are not followed.[18]Linux startup limits No effective activation or override resolution, recursive script execution, binary journal analysis or inspection of other users’ homes; most unit, cron and network text is reference evidence.[18] macOS startup artifacts Standard user/system LaunchAgent and LaunchDaemon XML/binary property lists, including Apple App cryptex folders; literal executable paths and process names are compared with indicators. Configuration does not establish execution; managed login items, unified logs and memory are excluded.[18] AndroidQF evidence Apps, process inventories, settings, files, logs, crashes, installation history and SMS/MMS backups. Collection permissions and Android versions constrain what is available; archives must first be extracted.[18] Android process interpretation JSON processes require a positive integer pidand acommand_linestring array; only the first command names the app. Olderprocesses.txtandps.txtinventories are also supported.[18]Android messages and certificates Only inbox messages are graded as received; sent, draft and failed messages are references. MMS attachments are not opened; encrypted SMS needs a password, and legacy password encodings may be unsupported. SHA-256 and SHA-1 certificate coverage gaps are counted separately.[18] iPhone evidence Local backups, independently encrypted artifacts and WALs, extracted filesystems, and selected sysdiagnose TAR/directory artifacts, including messages, profiles, browser/process records, running-process lists, installation logs, metadata and diagnostics. iOS 15, 16, 17 and 18 table layouts appear in the test suite.[18] iPhone exclusions and gaps No live/cloud acquisition, deleted-record recovery or unified-log decoding. Unsupported schemas, encoded-message variants, damaged WALs, encrypted root-manifest WALs, unavailable protection classes and record caps are gaps; source hashes do not authenticate the device.[18] iPhone acquisition boundary The readers do not unlock phones, create backups, download from iCloud or run MVT.[17] Optional Candiru patterns --patternsapplies two unchanged published rules: Microsoft’s 2021 DevilsTongue hijack-DLL rule and Malpedia’s automatically generated DevilsTongue rule. Matches are leads, not proof of infection, execution or coverage of every Candiru build.[16]Pattern-mode exclusions It does not acquire or scan running memory, decrypt payloads, inspect phone backups or add Pegasus detection.[16] Live connections Version 2 advertises listing current program connections and marking addresses on spyware lists without looking up or contacting those addresses.[20] Out-of-date programs Version 2 advertises local comparison of browsers, Office, Adobe Reader, Java, archive tools and other software against CISA exploited-vulnerability information and official fixed-version records; installed-program lists stay on the computer.[20] iPhone management profiles Version 2 advertises listing management, traffic-trust, VPN, proxy and DNS profiles with name, issuer and date.[20] Cellebrite traces Advertised checks examine USB hardware identities on Android and crash/computer identifiers on iPhone; a clean result cannot prove that no access occurred.[20] Coverage documentation The coverage catalogue is not a measurement of real-world detection or false-positive rates. --describeprovides the check catalogue, and reports list checks actually run and input-specific gaps.[18]Evidence grades and verdict interpretation
Grade or result Meaning on-diskA file has the fingerprint of a published spyware sample.[19] on-deviceA researcher-named app, signing certificate, file or setting is present on the phone.[19] file-patternA published detection rule matched a file; this is a lead, not proof.[19] persistenceThe name is registered to start automatically.[19] ran-commandA command or script naming it ran; it could be a researcher’s own command.[19] execution-recordA historical record names an executable without establishing its contents or resolved path.[19] inventory-recordA historical file inventory names it without establishing execution or current presence.[19] contactedEvidence demonstrates communication, such as a cookie, secure-connection record or download.[19] device-logThe phone’s own system logs name it.[19] receivedAn incoming SMS/MMS, iMessage or WhatsApp message contains it; opening or following it is not established.[19] dns-recordA name lookup was recorded, but contact is unverified.[19] attemptedSomething tried to reach it and was blocked.[19] configuredIt appears in local configuration.[19] referencedIt appears in saved content, which usually does not establish contact.[19] CLI exit 0No actionable matches and no recorded coverage gaps.[2] CLI exit 1Setup, argument or output error.[2] CLI exit 2A match beyond a saved reference.[2] CLI exit 3No actionable matches, but incomplete checks or indicator sources.[2] CLI exit 4The start menu closed without scanning; no verdict or report exists.[2] GUI exit Closing the window always returns 0, regardless of its displayed result; automation should use terminal commands.[2]Privacy, acquisition and release verification
Operation or issue Documented behavior Offline scan Python socket calls are refused, redirected files and Windows remote drives are rejected, and indicator addresses are neither resolved nor visited. POSIX network mounts and concurrent filesystem changes remain limitations.[11] Explicit updates --updatedownloads indicator lists from MVT, Amnesty International, Echap and Citizen Lab sources on GitHub, plus CISA’s exploited-vulnerability list and selected official CVE records. IP addresses reach GitHub and CISA, but installed-program lists are not sent.[11]Android acquisition --collect-android DIRis separate from offline scanning: after explicit owner consent it downloads a pinned official AndroidQF binary, verifies SHA-256 and launches it. Acquisition communicates with the phone, changes its state and saves local records.[11]Android setup The guide requires developer options, USB debugging, a data-capable USB cable, an unlocked phone and acceptance of the USB-debugging permission prompt.[14] Linux account selection The website advises running without sudo; using it may cause the check to inspect root’s browser records instead of the user’s.[14]Full reports Local reports can include usernames, browser profile names, paths, browsing details, SMS senders and excerpts. A cloud-sync client can independently upload output placed in a synced folder.[11] Minimal sharing report --share-json PATHincludes counts, scope labels, verdict and fixed advice, omitting indicator values, artifact text, paths, profile names and timestamps.[11]Temporary snapshots They can contain complete browser database bytes, including unqueried data. Interrupted runs or failed cleanup can leave copies in the system temporary directory; standalone scanning also writes temporary embedded indicator files.[11] Verification mechanisms Documentation distinguishes signatures, checksum checks, single-file Python rebuild comparison and desktop-program attestations. PyInstaller output is not byte-reproducible.[4] Verification limits A signature establishes who published the exact bytes, not absence of bugs; an older signed release need not be latest, and a checksum alone does not authenticate its publisher.[4] Security reporting The preferred route is a private GitHub security advisory; project Telegram is the stated alternative.[21] Disclaimed organizational relationships The NOTICE names Citizen Lab, Confiant, Recorded Future/Insikt Group, Amnesty International Security Lab and Access Now as organizations with which the project is not affiliated, endorsed, reviewed or connected.[13] Reviews, public scrutiny and release evidence
Source or release What the evidence establishes Agentbeam, v0.1.7 The reviewer cloned the repository and read the README, privacy, methodology, coverage documentation and network guard; no compromised-device detection test was performed.[5] Advertised “12,100 traces” Agentbeam could not verify the number and reported about 658 bundled indicator lines, with the remainder dependent on downloaded update feeds.[5] Elite Performance Exclusive, v0.1.10 The downloaded release matched the published checksum; the review examined network behavior rather than all code.[6][6] Elite’s maturity warning Its review described the tool at that time as three days old, maintained mainly by one person and changing several times daily; this is the review’s historical assessment.[6] DEV review, v0.1.8 Prodbymarcu read runtime modules and used --describe, reporting schema version 1.6 and tool version 0.1.8; the author stopped at preview rather than completing a full host scan.[7][7]GitHub issue #7, v0.1.5 The report describes crafted evidence capable of stalling or hanging a scan and characterizes the problems as availability issues, not code execution.[22] GitHub issue #14 The maintainer acknowledged an Android acquisition bug involving the collector’s ADB location and launcher path, and said it was fixed in v0.1.7.[23] GeekNews An xguru submission summarizes checking known commercial-spyware traces against published indicators of compromise.[8] Hacker News The located submission is titled “Zero Breach,” links interplains/zero-breachand shows two points; it concerns the separate base project.[10]Version 2 announcement Advertised additions include a 27-language GUI, Quick/Standard/Full scopes, progress and Stop controls, next-step guidance, phone checks, Cellebrite traces, live connections, outdated-program checks, iPhone profiles, about 900 new indicators and trainer practice mode.[24] v2.0.0 release The release identifies fixes since v0.1.15 and describes live connections, outdated-program comparisons, iPhone profiles and Cellebrite-trace checks.[20][20][20] v2.0.3 changelog entry An entry labelled “unreleased” warns that v2.0.0–v2.0.2 could show green for unreadable phone evidence, including empty folders, damaged sysdiagnose evidence and empty Android app lists; it changes such cases to “This phone could not be checked.”[25] v2.0.4 release The official release page identifies version 2.0.4 and directs v2.0.0–v2.0.2 users to update and recheck previously green phone results, stating that v2.0.3 fixed overstated checks.[12][12] -
Timeline
- October 27, 2026 — First toolkit version: The changelog dates v0.1.0 to this day and describes it as the first Zero Breach Toolkit version, built on Zero Breach; the corresponding GitHub release is dated October 27.[25][26]
- November 1, 2026 — GeekNews coverage: An xguru submission summarized the toolkit’s use of published indicators to check known commercial-spyware traces.[8]
-
Questions
What is Zero Breach Toolkit?
It is a free, open-source checker for known mercenary-spyware traces on computers, Android evidence and iPhone backups. It only finds published indicators in supported evidence and does not replace a forensic examination.[1][1]
What is the connection between Zero Breach Toolkit and Interplains?
interplainsis the GitHub username hosting the toolkit repository, and the profile displays ArchiveGenocide. The README credits archivegenocide, while verification documentation says the project has no named, publicly identified maintainers and approvers.[1][3][2][4]Where is Zero Breach Toolkit on GitHub?
The official repository is
interplains/Zero-Breach-Toolkit. Its entry point isscan/zero-breach-toolkit.py, with scanner implementation underscan/zbtriage/.[1]Has Zero Breach Toolkit had a security audit?
The available evidence does not establish a comprehensive independent security audit: the identified reviews cover selected code, documentation, network behavior and runtime preview. Public issue reports demonstrate scrutiny of specific failures, but not measured detection effectiveness.[5][6][7][7][22][5]
What do Zero Breach Toolkit reviews show?
Agentbeam, Elite Performance Exclusive and prodbymarcu provide limited inspections of early versions rather than full compromised-device testing. Agentbeam also could not verify the advertised “12,100 traces” figure, reporting about 658 bundled indicator lines and additional coverage dependent on downloaded feeds.[5][6][7][7][5]
How does Zero Breach Toolkit work as a spyware scanner?
It examines supported browser, file, operating-system and local phone artifacts for published indicators, distinguishing exact fingerprints from execution records, contact evidence and saved references. Missing evidence and unpublished or expired indicators limit the meaning of a clean result.[18][18][18][19][1]
-
Caveats
- Unsigned-app wording: The README says Microsoft and Apple signing has not happened “yet,” while verification documentation says no signed Windows release is planned while the project remains anonymous. Both agree that the native programs are unsigned; their wording differs on future expectations.[2][4]
- Changelog versus release status: The changelog labels its v2.0.3 entry “unreleased,” while the v2.0.4 official release states that v2.0.3 fixed the phone-result problem and presents v2.0.4 as a release. The official release record therefore supplies publication evidence that the changelog label does not.[25][12][12]
- Promotional totals versus verified scope: Agentbeam could not verify the “12,100 traces” total and counted about 658 bundled indicator lines; additional indicators depended on downloaded feeds. Neither that total nor the project’s coverage catalogue establishes a real-world detection or false-positive rate.[5][18]
- “Read-only” and “offline” are bounded claims: The documented scan preserves source evidence and system settings but writes temporary snapshots and requested reports. Its network guard is not an operating-system sandbox, and Android acquisition and explicit updates operate outside the offline-scan boundary.[2][11][11][11]
- Review claims have limited scope: Elite Performance Exclusive explicitly limited its inspection to network behavior, and the DEV reviewer stopped before a full host scan. These reviews should not be treated as comprehensive certification of the toolkit.[6][7]
-
Sources
- [1]GitHub - interplains/Zero-Breach-Toolkit: Free, open-source check for traces of mercenary spyware (Pegasus, Predator, Graphite, Candiru and more) on computers, Android phones and iPhone backups. Double-click apps for Windows, Mac and Linux, in 27 languages. · GitHub github.com
- [2]Zero Breach Toolkit raw.githubusercontent.com
- [3]interplains (ArchiveGenocide) · GitHub github.com
- [4]Verifying a release raw.githubusercontent.com
- [5]Zero Breach Toolkit: What the Free Spyware Scanner Does, and How to Vet It Before You Run It agentbeam.com
- [6]Zero Breach Toolkit: Is the New GitHub Spyware Checker Real and Safe to Run? eliteperformanceexclusive.com
- [7]What does a clean scan actually prove? dev.to
- [8]Zero Breach Toolkit - PC와 스마트폰의 스파이웨어 흔적을 오프라인으로 검 | GeekNews news.hada.io
- [9]GitHub - interplains/zero-breach: Zero Breach — local, read-only scanner that checks a computer for indicators of the AdFection / Candiru ad-delivered spyware campaign. No network, no admin, no dependencies. · GitHub github.com
- [10]Zero Breach | Hacker News news.ycombinator.com
- [11]Privacy raw.githubusercontent.com
- [12]Release Zero Breach Toolkit 2.0.4 · interplains/Zero-Breach-Toolkit · GitHub github.com
- [13]raw.githubusercontent.com/interplains/Zero-Breach-Toolkit/main/NOTICE raw.githubusercontent.com
- [14]Free, open-source, offline checks for traces of known mercenary spyware. Works on Windows, macOS and Linux, Android phones and iPhone backups. zerobreachtoolkit.com
- [15]Zero Breach Toolkit sourceforge.net
- [16]Optional Candiru file patterns raw.githubusercontent.com
- [17]Local iPhone evidence raw.githubusercontent.com
- [18]Coverage raw.githubusercontent.com
- [19]Methodology raw.githubusercontent.com
- [20]Release Zero Breach Toolkit 2.0.0 · interplains/Zero-Breach-Toolkit · GitHub github.com
- [21]Security policy raw.githubusercontent.com
- [22]Crafted evidence can stall or hang a scan (matcher tokens, .ips regex, hashscan FIFO) · Issue #7 · interplains/Zero-Breach-Toolkit github.com
- [23][Windows] Wrapper fails to detect downloaded ADB executable during Android acquisition · Issue #14 · interplains/Zero-Breach-Toolkit github.com
- [24]Version 2 is out · Issue #32 · interplains/Zero-Breach-Toolkit github.com
- [25]Changelog raw.githubusercontent.com
- [26]Release Zero Breach toolkit v0.1.0 · interplains/Zero-Breach-Toolkit · GitHub github.com