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>
|
||
|---|---|---|
| assets/icons | ||
| crates | ||
| dist | ||
| gui | ||
| packaging | ||
| .editorconfig | ||
| .env.example | ||
| .gitignore | ||
| Cargo.lock | ||
| Cargo.toml | ||
| README.md | ||
| rust-toolchain.toml | ||
Hound Antivirus
A premium, freemium antivirus for Linux. One engine, three faces:
a Rust daemon (houndd), a CLI (hound), and a Tauri GUI with a
system-tray sentinel that changes color with your security state.
Built for the distros people actually run: Ubuntu, Debian, Linux Mint, and anything else that ships ClamAV.
Repository layout
antivirus/
├── Cargo.toml # Rust workspace
├── crates/
│ ├── hound-api/ # shared wire types + socket client (daemon/CLI/GUI all use it)
│ ├── houndd/ # the daemon: Unix-socket API over a pluggable engine
│ └── hound/ # CLI client
├── assets/icons/ # dog-head brand mark + 4-state tray ladder
└── gui/ # Tauri 2 desktop app (system tray + scan UI)
Architecture
houndd (Rust daemon — the engine)
┌──────────────────────────────────┐
│ ScanEngine trait │
│ ├─ L1 ClamAV signatures (now) │
│ ├─ L2 Curated threat packs(Pro)│
│ ├─ L3 Behavioral monitor (Pro) │
│ └─ L4 Supply-chain checks(Pro) │
└──────────────┬───────────────────┘
Unix socket (JSON-RPC, line-delimited)
┌───────────┼───────────┐
hound CLI GUI (Tauri) future modules
The daemon is the only process that touches a scanning engine. CLI and GUI are thin clients — so future suite tools (firewall, updater, …) plug into the same socket.
Swapping the engine (the ClamAV seam)
ClamAV is a temporary dependency. Everything ClamAV-specific — version
probe, signature freshness, the clamscan subprocess + output parsing,
freshclam — lives in one file behind a four-method trait:
crates/houndd/src/engine.rs
trait ScanEngine { name; probe; scan; update }
struct ClamAvEngine // today
const ENGINE: ClamAvEngine // ← flip this line when the native
// engine lands; nothing else in the
// daemon, CLI, GUI, or wire API moves
The wire stays engine-agnostic: Status.engine names the implementation
("clamav" today) and Status.db carries what any signature store has —
a file name and a timestamp. When our own Rust engine ships, it's a new
ScanEngine implementation, a one-const flip, and the tray/CLI/GUI simply
start reporting the new engine name.
Icon system
The brand mark is a solid dog head (assets/icons/hound.svg), a single
flat fill. It ships in two treatments:
- Brand ladder
hound-{16,22,24,32,48,256}.png— native periwinkle#9896E0, for the window icon, About box, and marketing. - Tray-state ladder
state-<name>-{16,22,24,32,48}.png— the same path re-tinted per security state, for the system-tray sentinel:
| State | Fill | Meaning |
|---|---|---|
protected |
#22C55E green |
up-to-date / protected / clean |
scanning |
#F59E0B amber |
scan in progress / signatures need update |
threat |
#EF4444 red |
infection found |
paused |
#6B7280 gray |
real-time monitor off |
Green is the good state; amber is work in progress, never a failure.
Quickstart (development)
Prereqs: Rust (see rust-toolchain.toml), Node 20+, ClamAV, and the
Tauri system libs (libwebkit2gtk-4.1-dev, libgtk-3-dev, libayatana-appindicator3-dev).
# 1. Signatures (needs the clamav freshclam DB)
sudo freshclam
# 2. Daemon (terminal 1)
cargo run -p houndd
# 3. Scan (terminal 2)
cargo run -p hound -- status
cargo run -p hound -- scan ~/Downloads
# 4. GUI
cd gui && npm install && npm run tauri dev
Verifying the engine with the EICAR test file
EICAR is the industry-standard 68-byte test signature — every AV that works will flag it. Generate it and scan it:
printf 'X5O!P%%@AP[4\\PZX54(P^)7CC)7}$EICAR-STANDARD-ANTIVIRUS-TEST-FILE!$H+H*' > /tmp/eicar.com
cargo run -p hound -- scan /tmp/eicar.com
# expect: exit code 1, "Eicar-Test-Signature FOUND"
Updating signatures
hound update wraps freshclam (trying sudo freshclam first, since
plain-user runs can't write /var/lib/clamav and /var/log/clamav). The
GUI's "Update Signatures" button drives the same RPC and shows the log.
cargo run -p hound -- update # or: hound update --json
The GUI (gui/)
A Tauri 2 desktop app — a thin view over the same houndd socket the
CLI uses (via the shared hound-api client), so the window and the
command line never disagree about your security state.
gui/
├── dist/ # the front-end (vanilla HTML/CSS/JS, premium dark shell)
└── src-tauri/ # Tauri 2 shell + system-tray sentinel
The tray sentinel swaps the 4-state icons (green/amber/red/gray) as your state changes; the window shows a live protection hero, a scan progress bar, a results table, and the signature-update log.
Build it:
cd gui
npm install
npm run tauri dev # dev with hot reload
npm run tauri build # → .deb in src-tauri/target/release/bundle/
Git conventions
-
Branch
mainis deployable; small, focused commits. -
No hardcoded secrets. For pushes, the bot token lives in a repo-local credential file (never tracked):
git config credential.helper 'store --file=.git/.git-credentials' chmod 600 .git/.git-credentials echo 'https://<user>:<token>@git.joelovestech.com' >> .git/.git-credentials -
Commits: imperative subject, ≤ 72 chars. Example:
houndd: add line-delimited JSON-RPC socket API
License
TBD — core daemon likely proprietary (freemium), shared CLI possibly OSS.
Decision pending; workspace.package.license = MIT is a placeholder.