Repository navigation
Releases
Releases · SilentMesh-Systems/Lure
v0.1.0b3
- Anonymous usage statistics (command, version, OS, alert severity and service only; never IPs, bait names or alert contents), on by default and stated on first run; off with LURE_NO_TELEMETRY=1 or DO_NOT_TRACK=1.
- Weekly "still listening" message, plus one at start, to your alert channels;
--heartbeat-dayssets the interval, 0 turns it off. - Bait email uses your organisation's domain (
lure initasks;--email-domainon generate) instead of the giveaway<profile>-it.local; rotate keeps it. lure initreminds you to tell IT before planting baits on shared drives.
Fixes from the founder's own test on Windows, 15 Nov 2026.- Dashboard rebuilt as an operator console in the SilentMesh console design: an incident queue with one incident per source, a detail pane (why it was flagged, timeline with ATT&CK technique, impact including what Lure cannot see, what to do, handling), a 24-hour histogram, live feed, bait inventory, integrations, CSV and JSON export, keyboard shortcuts, and Run test that fires a real bait login at your own listener. Fonts are bundled, so the dashboard makes no outside requests.
- Every alert now says who (this computer, a device on your network, or the internet), what bait, how it was used, and what to do, in the dashboard, webhooks, email and popups.
- Demo alerts are marked as demo everywhere, with a banner on the dashboard, so a demo can never look like a real attack.
- Alerts without a webhook:
--desktoppopups on Windows, Linux and macOS,--email-tothrough your own Gmail or Outlook (app password from LURE_SMTP_PASSWORD, never saved), and readable ntfy push messages. lure initasks two questions: home or work computer, and how you want to be told. The choice is saved outside the bait folder.- No real company names in baits: vendors are roles such as "AMC vendor" or "automation vendor" in every profile.
- Home profile writes a plain
passwords.txtstyle note and a Wi-Fi QR, not an office spreadsheet. - Windows: commands print as
py -m lure; if port 3389 is taken by Remote Desktop, the RDP bait uses 13389, and listen, test and rotate read ports from the manifest. - Updates by approval:
lure listenchecks PyPI once a day and only tells you;lure updateshows what changed and installs only when you type yes. Off with --no-update-check or LURE_NO_UPDATE_CHECK=1. Releases publish only after tests pass, the tag matches the version, and a reviewer approves.
Install: pipx install --pip-args=--pre silentmesh-lure
v0.1.0b2
MCP spec 2026-07-28, lure test, lure init, three-command start. pip install silentmesh-lure
v0.1.0-beta
Lure by SilentMesh v0.1.0-beta
Planted credentials that give the attacker away. A quiet safety net for any network, from a home router to a data centre. Free and open source (Apache-2.0).
What it does
lure generate writes eight believable bait files with fake logins:
- a messy password workbook
- a saved Remote Desktop file
- a vendor VPN profile
- a browser password export
- a
.git-credentialsfile - a leaky email
They point at each other the way real ones do. lure listen waits on a spare machine. When anyone tries a bait login, you get one alert naming the file, where you planted it, and the path the attacker took.
Highlights
- Alerts on use, not on open: antivirus scans, backups and indexing never page you.
- Nothing is sent to us: alerts go to your own webhook, syslog, terminal or dashboard.
- Local dashboard (
lure listen --ui), with bait status, a live alert feed and the attacker's path. - MCP server (
lure mcp) with 7 tools, so AI agents can check, rotate and triage. It never returns passwords. - Profiles:
it(default),office,home,bank,hospitalandplant. lure rotate: fresh baits; old ones keep alerting, which catches credentials stolen earlier.lure record: stores exactly where each bait was planted.lure watch: catches bait usernames typed into your real systems, from their logs (SSH, VPN, RADIUS, web, Windows 4625/4771).- DNS baits (
--dns-zone): a bait hostname that alerts on lookup, through almost any firewall. - Alerts formatted for Slack, Teams, Discord and Telegram.
- QR-code and bookmark baits: scanning the QR or opening the bookmarked console raises an alert.
- Bring your own profile as a JSON file; legal, education and retail examples included.
- Runs on Python 3.9 and newer, standard library only. Tested on Linux (x86-64 and ARM, including Raspberry Pi) and Windows.
Verified before release
- Automated tests: 30 tests pass on CI for Linux x86-64, Linux ARM and Windows.
- Real-client tests:
- real
gitsends the.git-credentialsbait - a real RDP client (FreeRDP 3) uses the RDP bait
- the HTTP baits work with curl and browsers
- the official MCP Python SDK client connects to the MCP server
diglookups of a DNS bait, and an SSH log line with a bait username, both raise high alerts
- real
Known limits
- Real-system detection depends on the logs you point
lure watchat. - Not yet: cloud bait keys, QR-code baits, a hosted demo listener.
- macOS is not yet supported.
- This is a beta, supported on a best-effort basis.
Install
git clone https://github.com/SilentMesh-Systems/Lure && cd Lure
python3 -m lure generate --listener <listener IP> --out baits
python3 -m lure listen --manifest baits/manifest.json --ui