Closes Phase 3's delivery half. `hound update` now asks the definitions
host what exists, downloads what this machine lacks, verifies it, and
installs it. Verified end to end against the live host over TLS:
starting from an empty directory:
defs: no verified packs were found
hound update:
definitions: 1 pack(s) installed, 0 already current
crates-io-2026.08.21.pack — 19 indicators, version 2026.08.21
loaded 19 indicators from 1 pack(s) [2026.08.21]
and immediately afterwards a lockfile naming rustdecimal is flagged,
while tokio beside it is not.
running it again:
definitions: definitions are up to date (1 pack(s))
a pack with one byte flipped in its signed payload, advertised in the
index with a correct hash and a version of 9999.1.1:
definitions: 0 pack(s) installed, 1 up to date, 1 REJECTED
tampered-test.pack: the definitions pack is not signed by Hound and
was discarded
Three refusals are the design:
* THE INDEX IS A HINT, NEVER AN AUTHORITY. It says which packs exist and
what they hash to, and both are unverified — anyone who can serve the
index can lie about either. Only the Ed25519 signature decides whether
a pack is real. A tampered index can waste bandwidth and nothing else,
which is exactly what the test above demonstrates: correct hash,
higher version, still refused.
* NOTHING UNVERIFIED REACHES THE DEFINITIONS DIRECTORY. Downloaded to a
temp file, verified there, then moved with a rename inside the same
directory so it is atomic. The daemon cannot observe a half-written
pack, and a crash mid-download leaves a stray temp file rather than a
loadable one.
* A FILENAME FROM A REMOTE INDEX IS UNTRUSTED INPUT. The updater runs as
root, so an entry of ../../../etc/cron.d/evil.pack would be remote code
execution. Only a plain basename ending in .pack, with no separators,
no dot-dot and no leading dot, is accepted. Tested against eight
hostile shapes.
Sizes and timeouts are bounded — a pack is a list of package names, so a
server offering a hundred gigabytes is broken or hostile and the
difference does not matter to a full disk. The read is bounded
independently of Content-Length, which is also just something the server
said.
Definitions are fetched BEFORE rules are reloaded, and a failure to fetch
does not stop the reload. No network, a mirror down, a pack that will not
verify — none of those are a reason to skip the half that still works.
HOUNDD_DEFS_URL points the client at a mirror, which matters for the
air-gapped deployments that are a real part of the Fleet story.
build-pack now writes index.json beside the pack, so publishing is one
command.
358 tests pass.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Four tools over MCP stdio, so a coding assistant can ask Hound about a
project at the moment it matters rather than after:
hound_check_project the full supply-chain sweep
hound_check_package is this package known bad, before installing it
hound_check_file one file: a model, a lockfile, a manifest
hound_check_mcp_config audit the servers your assistant already trusts
READ-ONLY, permanently. There is no hound_quarantine, no hound_delete,
no way to change a setting. An agent can be persuaded by the very
repository it is inspecting — that is the threat this product exists to
detect — so a destructive MCP tool would hand the attacker exactly the
capability they were reaching for. A test asserts every tool name starts
with hound_check_ and contains none of quarantine/delete/remove/restore/
settings/set/update/install/write/exec, so the property cannot erode.
The output is written to be read twice: by the model that called the
tool, and by the human reading that model's summary. That rules out rule
identifiers and jargon — a test greps the output for both — and it rules
out ambiguity about severity. A model reading "1 warning" may well decide
to proceed, so a critical finding leads with an explicit recommendation
and names the consequence: this runs code during installation, before
any of the project's own code runs.
Clean results say what they did NOT check. "Nothing wrong" that reads as
a blanket endorsement is worse than no answer, so a clean project notes
it is not a source review, a clean name check notes it did not read the
package's contents, and a clean MCP audit still points out that every
server listed runs with your permissions and is called without asking.
The server passes its own audit, which was the point:
{ "mcpServers": { "hound": { "command": "/usr/bin/hound-mcp" } } }
No npx, so nothing is fetched at launch. No env, so no secret is handed
over. No path argument, so it is granted no directory. Hound's MCP audit
flags all three in other people's configs; a security tool that failed
its own check would have answered the only question that mattered.
Verified against the installed binary.
Protocol notes, since both are easy to get wrong and fatal:
- a notification carries no id and must never be answered; MCP sends
notifications/initialized straight after the handshake, so replying
corrupts the stream on the first exchange
- an id of 0 is still an id, and treating it as absent silently drops
the first call from any client that counts from zero
- a tool failure is a RESULT with isError, not a JSON-RPC error: the
agent should see "I could not read that path" as an answer it can act
on, not a transport fault that looks like a broken server
- nothing but protocol messages ever goes to stdout; diagnostics go to
stderr, because one stray println corrupts the session
Shipped in the .deb, and the postinstall now prints the config entry.
334 tests pass.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Lockfile parsing for npm (all three lockfile versions), yarn, cargo,
poetry, requirements.txt, go.sum, Gemfile.lock and composer.lock, wired
through the indicator index so a sweep checks real dependencies against
real definitions. Signed packs load in the daemon; the sweep gets the
index; `hound supply-chain` cites the OSV record it matched.
A lockfile is the right thing to read: it names every transitive
dependency at an exact version in one small file, and it lists what WILL
be installed rather than what already is — which matters when the
payload runs during installation.
Every parser is hand-written rather than pulling in a TOML and a YAML
crate. Two fields from each format, and a scanner parsing hostile input
should have as little parsing surface as it can.
The important part of this commit is a false positive it fixes.
Building a pack from the whole crates.io OSV export and sweeping a
project produced TWO criticals: rustdecimal, correctly, and **tokio
1.38.0**, which is not malware and never has been. The export is 1,524
GHSA and 1,206 RUSTSEC vulnerability advisories against 19 malicious-
package records, and the parser treated all of them as malware.
GHSA-2grh-hm3w-w7hv describes a tokio race condition fixed in 1.8.1;
Hound reported a version released years later as malicious.
Two independent bugs, either of which alone is fatal:
* Vulnerability advisories were ingested at all. A malicious package
should not exist; a vulnerable one is a legitimate library with a bug
and most of its versions are fine. Records must now PROVE they are
malicious-package reports — a MAL- id, the malicious-packages-origins
marker, or GHSA's "Malicious code in" wording — and anything
unrecognised is dropped.
* Unrecognised version ranges fell back to "all versions", which is the
opposite of safe. That is what turned a range of 1.8.0-to-1.8.1 into
a verdict on every tokio ever published.
Rebuilt against the same input, the pack now holds 19 indicators rather
than 3,614, rustdecimal is still caught and cites MAL-2022-1 rather than
a GHSA advisory, and tokio and serde are clean. The real tokio advisory
is now a regression fixture, because anything that flags tokio is a
product nobody trusts twice.
Also: definitions loading fails CLOSED on authenticity and OPEN on
everything else. No trusted key means no definitions and a message
saying so, because an operator who believes they are protected and is
not is worse off than one who knows. A pack that fails verification is
skipped and the rest still load. No packs at all is a working daemon —
install scripts, prompt injection, pickles and MCP audits need no feed.
There is deliberately no placeholder signing key compiled in. A fake key
that looks real is how a development shortcut becomes a shipped
vulnerability; an empty trust store is noisy in the way that gets fixed
before release. HOUNDD_DEFS_KEY supplies one for development.
294 tests pass across the workspace.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Phase 3's foundation. Three jobs that share a data model:
osv parse the ossf/malicious-packages feed (Apache-2.0, ~226k
records, daily) into indicators
index answer "is this package known bad?" fast enough to ask it
thousands of times per project sweep
pack sign a definitions pack, and verify one before loading it
Validated against the real feed rather than fixtures: the whole
crates.io OSV export, 2,749 records, parsed with zero failures — 3,885
indicators across ten ecosystems, 19 of them MAL-. `rustdecimal` (the
real typosquat of rust_decimal) resolves in crates.io and stays clean
in npm and PyPI, which is the ecosystem isolation working.
Notes on the three:
* A malicious-package record is not a vulnerability record. It almost
always carries introduced:"0" with no fix, meaning EVERY version is
malicious — the package exists only to be malware, so there is no safe
version to upgrade to. Conflating that with a version-bounded
vulnerability either misses real hits or condemns safe versions of
legitimate packages, so the two are modelled separately.
* The index is a cuckoo filter in front of a map. Cuckoo rather than
bloom specifically because a definitions feed needs DELETION: OSV
withdraws records — it once withdrew 157 malware reports after a
false-positive incident — and a filter you cannot remove from means a
withdrawn record costs a probe forever or forces a rebuild. 226k
indicators fit in under 4 MB; the crates.io set is 8 KB.
The property that must never break is no false negatives, and it has
its own test. A false positive costs a hash lookup; a false negative
is malware reported as clean. That is also why a fingerprint hashing
to zero is nudged to one — zero marks an empty slot, so without the
nudge one key in 65,536 would silently vanish.
* Signing is not about entitlement; the subscription gates the server.
It answers "is this pack really from us?", because someone who can
substitute a definitions file can add an entry for /usr/bin/sudo and
have Hound quarantine it on every machine that updates — a supply
chain attack delivered through the security product. Verification
happens on the raw bytes BEFORE anything parses them, so a hostile
pack never reaches the parser at all.
256 tests pass across the workspace.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Completes Phase 5. Half of a Linux compromise is not a file on disk, it
is a line added to a startup file — a curl in a shell profile, a systemd
unit with a dull name, one extra key in authorized_keys. The payload is
often unremarkable; what makes it an incident is that it survives a
reboot and nobody reads those files from one year to the next.
So this is not a scanner but an inventory with a memory. It records
systemd units (system and per-user), cron in all its locations,
autostart entries, shell profiles, authorized_keys and ld.so.preload,
then reports what CHANGED.
Three decisions, all of which are the difference between a report people
read and one they turn off:
* Content is hashed, not stat'd. An mtime can be set backwards with one
touch, and someone editing a startup file is exactly the person who
would. Verified: a backdated edit is still caught.
* A first run reports no changes and says so. Everything would be a
change, and a first-run report full of alarms is one nobody reads.
What a first run can honestly say is how many entries no package
claims — 97 of 1,026 on this machine — because that is true
regardless of history.
* Writing the baseline is an explicit act (`--accept`, or
update_baseline on the wire). A plain check must never quietly record
whatever is currently installed as normal; that is how a compromise
becomes the new baseline.
Package ownership decides what is ordinary: a unit that arrived with a
package is the system working, the same unit unowned is somebody's
decision. Reuses the merged-/usr-aware index from the rootkit rewrite,
with a test asserting most units resolve to a package — if that ratio
collapses, ownership lookup has broken and the whole report is noise.
Exercised end to end against this machine: baseline of 1,026 items,
a planted user unit caught as ADDED, an in-place edit with a backdated
mtime caught as CHANGED, and its deletion caught as REMOVED. The test
artifact was removed afterwards.
Cross-distro verification of the rootkit rewrite is now MET. Henry ran
the suite on the Ubuntu box (26.04, glibc 2.43): 20/20, including both
unowned_setuid_does_not_fire_on_a_healthy_system and
no_false_positives_on_system_binaries.
214 tests pass across the workspace.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Phase 4's detection core, as a standalone crate. This is the part with no
competitor on Linux, and deliberately the part with no Linux in it —
file parsing and logic only, no fanotify, no /proc, no eBPF — so the
macOS and Windows port is weeks rather than a second product.
Five detectors, 88 tests:
pickle GLOBAL/STACK_GLOBAL walk over .pt/.ckpt/.pkl/.joblib.
torch.load runs a stack machine; a model file is a
program and downloading weights is a code-execution
decision.
injection Instructions aimed at a coding agent in CLAUDE.md,
AGENTS.md, .cursorrules, copilot-instructions.
installscript preinstall/postinstall hooks that curl|sh, decode and
run, reach for credentials, or install persistence.
typosquat Damerau-Levenshtein against popular names, plus
slopsquat detection: new + near-zero downloads + one
edit from something popular is the signature of a name
a model invented and somebody then registered.
mcp Servers fetched unpinned at launch, handed secrets, or
pointed at $HOME or credential paths.
Wired through `supply.sweep` on the socket and `hound supply-chain
<path>`, which exits 1 on a critical so it drops into CI.
Three things worth recording:
* Scoring is by independent category, not by keyword count. One
suspicious phrase is a phrase; two categories at once is an attack.
A file that only says "ignore previous instructions about formatting"
is a warning, not a critical.
* Proximity matters more than presence. The first version flagged an
entirely ordinary conventions file, because it mentioned ".env" in
one paragraph and "prefer small commits" in another. A credential and
a movement verb now have to appear within a sentence of each other.
The test that caught it is kept as the regression.
* Pickle call detection has to come out of the opcode walk, not a byte
search. REDUCE, INST and OBJ are the ASCII letters R, i and o, which
also occur inside every string the stream carries — searching raw
bytes finds the o in "os" and reports a call that never happens,
turning every warning into a critical.
Every finding carries a plain-language explanation and a next step, and
there is a test asserting explanations do not leak rule identifiers or
jargon. The audience includes people who cannot triage a YARA match and
should never be shown one.
Verified against a demo project holding a squatted @vue plugin with a
curl|sh postinstall, a poisoned CLAUDE.md in a vendored repo, an
unpinned MCP server holding a GitHub token, and a pickle calling
os.system — four criticals and one warning, while the legitimate
CLAUDE.md, the real express manifest and the properly-scoped MCP server
beside them stayed clean.
189 tests pass across the workspace.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
FAN_OPEN_EXEC_PERM hands us the open and waits for an answer, so a
binary can be refused before it runs. inotify could only report what
had already happened.
Verified end to end as root against a dedicated tmpfs (examples/
gate-smoke.rs, three phases):
benign binary ran 7.2 ms
EICAR binary blocked 1.8 ms never executed
scanner stalled 5 s ran 1.6 s watchdog rescued 3 events
The third phase is the one that matters. A gate that can hold a process
forever is a machine-wedging bug wearing a feature's clothes, so the
watchdog answers ALLOW for anything unanswered past DEADLINE and counts
it. A missed detection is a bad day; a frozen machine ends the product.
Two things this cost, both worth recording:
* Scanning by re-opening the path deadlocks the daemon against itself.
The open() lands on the watched mount and queues a permission event
behind the one we are currently answering, and we cannot answer that
one until we finish this one. Allowing our own pid does not help —
the thread never gets back to the queue to apply the rule. The gate
reads through the descriptor the kernel already handed it, with
pread so the gated process still sees its own file offset. This is
what hung the first smoke run.
* The watchdog can only rescue events it has been told about, and it
learns of them when the queue is drained. Scanning on the draining
thread makes every event behind a slow scan invisible to the
deadline. Reader and workers are therefore separate threads: the
reader never blocks on a scan, so every event is registered within
microseconds of arriving.
Also:
- ScanEngine::scan_bytes — the seam the gate needs, since it must never
scan by path. Engines that cannot do it return None and simply are
not usable behind the gate.
- Settings gain exec_gate and exec_gate_paths, defaulting to OFF. It
needs CAP_SYS_ADMIN and a root-filesystem mark holds every process on
the box; that is not a default to ship before Phase 2 soak testing.
- ABI constants are defined locally rather than taken from libc, so a
version bump cannot quietly change what we ask the kernel for.
78 tests pass, up from 57.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The old engine shelled out to clamscan for every scan, and clamscan
reloads a 169 MB signature database on every invocation. Measured on a
68-byte EICAR file: 6.5 seconds and ~1.5 GB RSS — paid once per file,
and realtime.rs called it once per inotify event.
Replaces it with HoundEngine: yara-x compiled once at daemon start,
held in memory, one scanner reused across a whole walk, plus a verdict
cache keyed on (dev, ino, mtime, size) so an unchanged file that has
been seen before never reaches the matcher.
Measured after, same machine, same EICAR file:
single file 6.5 s -> 4 ms
400 files cold -- -> 9 ms
400 files warm -- -> 5 ms
Also here:
- rules.rs: hot-swappable rule store. Built-in pack is embedded so a
fresh install detects something before it has ever reached the
network; on-disk packs load from $HOUNDD_RULES_DIR, /var/lib/hound
or the XDG data dir. Reload swaps an Arc, so in-flight scans are
never torn out from under.
- cache.rs: bounded FIFO verdict cache. Any of the four key fields
changing means rescan, so edits, truncates and replace-by-rename all
correctly miss.
- The goodware gate: every rule is scanned against all of /usr/bin,
/bin and /usr/sbin in CI, and a single hit fails the build. It has
already earned its keep — it caught a reverse-shell rule that matched
/usr/bin/sudo, which is now removed rather than tuned. A rule that
quarantines sudo is worse than no rule at all.
- ScanEngine is Send + Sync and selection stays per-call, so
HOUNDD_ENGINE=clamav still reaches the legacy path for comparison.
- ScanResult.skipped reports files passed over for size instead of
quietly counting them as clean.
- Settings gain theme (auto/light/dark), tray_icon_style (color/mono),
close_to_tray and confirm_quit, normalised daemon-side because
clients are not trusted to send a theme we can render.
57 tests pass, up from 29.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
- hound-api: DbFile (file + updated_at), Status.db, UpdateResult, Client::update()
- houndd: probe() extracts signature freshness from /var/lib/clamav mtimes;
update() tries 'sudo freshclam', falls back to plain 'freshclam' and
reports the honest reason on failure; output capped to 2KB tail
- hound CLI: 'hound update' with --json parity and human output
- rename Status.clamav_present -> engine_present + add Status.engine
(prep for the engine seam; wire stays engine-agnostic)
Live-verified: status shows DB file + last-modified RFC3339; update
returns ok:false with the real log for a plain user (exit 1).
Hound
