fips/CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

Added

• Linux release artifact workflow: builds x86_64 and aarch64 tarballs and .deb packages on v* tag push, with SHA-256 checksums
• AUR publish workflow for tagged stable releases

Security

• Bloom filter poisoning defense. Reject inbound FilterAnnounce messages whose false-positive rate exceeds a configurable cap (node.bloom.max_inbound_fpr, default 0.05). Previously a peer running a modified build could send an all-ones filter, causing (1) lookup attraction / black-hole routing for unknown targets, (2) aggregation contamination as the poisoned bits propagated one hop per announce tick via strict-OR merging, and (3) mesh-size estimate blowup to f64::INFINITY. Rejection is silent on the wire; rejected announces log at WARN and increment a new bloom.fill_exceeded counter. The peer's prior accepted filter and sequence number are preserved on rejection so a single bad announce cannot wipe a peer's contribution to aggregation. An independent self-plausibility WARN fires (rate-limited to once per 60s) if our own outgoing filter ever exceeds the cap, surfacing aggregation drift or ingress-check bypasses. BloomFilter::estimated_count now returns Option<f64> and returns None for saturated filters, preventing f64::INFINITY from propagating into mesh-size estimates. The node-level estimated_mesh_size field (already Option<u64>) propagates None when any contributing filter is above cap.

Fixed

• Control socket path detection in fipsctl and fipstop now checks for the /run/fips/ directory instead of the socket file inside it, so users not yet in the fips group get a clear "Permission denied" error instead of a misleading "No such file" fallback to $XDG_RUNTIME_DIR (#30, reported by @Sebastix)
• OpenWrt ipk build excluded BLE feature that requires D-Bus, which is unavailable on OpenWrt targets
• IPv6 routing policy rule added at TUN setup to protect fd00::/8 from interception by Tailscale's table 52 default route
• Bloom filter routing no longer swallows traffic when no bloom candidate is strictly closer than the current node. find_next_hop now falls through to greedy tree routing in that case instead of returning NoRoute, which previously caused dropped packets in topologies where the tree parent was closer but not a bloom candidate
• Auto-connect peers now reconnect after a graceful Disconnect notification from the remote side. handle_disconnect previously removed the peer without scheduling a reconnect, orphaning the entry on a clean upstream shutdown; the other removal paths (link-dead, decrypt failure, peer restart) already scheduled reconnect (#60, reported by @SwapMarket)
• fipsctl connect now rejects FIPS mesh (fd00::/8) addresses for udp, tcp, and ethernet transports with a clear error message instead of echoing success while the daemon silently failed the bind with EAFNOSUPPORT (#61, reported by @SwapMarket)
• Tighten TreeAnnounce ancestry validation to match the spanning tree specification. The receive path now verifies that the ancestry is structurally consistent with the signed parent declaration before mutating tree state.

[0.2.0] - 2026-03-22

Added

Operator Tooling

• fipsctl connect and disconnect commands for runtime peer management via control socket, with hostname resolution from /etc/fips/hosts

IPv6 Adapter

• Pre-seed identity cache from configured peer npubs at startup, so TUN packets can be dispatched immediately without waiting for handshake completion (@v0l)

Mesh Peer Transports

• New Tor transport with SOCKS5 and directory-mode onion service for anonymous inbound and outbound peering
• DNS hostname support in peer addresses for UDP and TCP transports
• Non-blocking transport connect for connection-oriented transports (TCP, Tor)

Packaging and Deployment

• Reproducible build infrastructure: Rust toolchain pinning via rust-toolchain.toml, SOURCE_DATE_EPOCH in CI and packaging scripts, deterministic archive timestamps
• Top-level packaging Makefile for unified build across formats
• Kubernetes sidecar deployment example with Nostr relay demo
• Nostr release publishing in OpenWrt package workflow
• SHA-256 hash output in CI build and OpenWrt workflows

Testing and CI

• Maelstrom chaos scenario with dynamic topology mutation and ephemeral node identities via connect/disconnect commands
• Consolidated Docker test harness infrastructure

Changed

• Discovery protocol: replace flooding with bloom-filter-guided tree routing. Includes originator retry (T=0/T=5s/T=10s), exponential backoff after timeouts and bloom misses, and transit-side per-target rate limiting. Removed 257-byte visited bloom filter from LookupRequest wire format. This is a breaking change; nodes running versions prior to this release will not be compatible.

Fixed

• DNS responder returned NXDOMAIN for A queries on valid .fips names, causing resolvers to give up without trying AAAA. Now returns NOERROR with empty answers for non-AAAA queries on resolvable names. (#9, reported by @alopatindev)
• Stale end-to-end session left in session table after peer removal blocked session re-establishment on reconnect — remove_active_peer now cleans up self.sessions and self.pending_tun_packets. (#5, @v0l)
• schedule_reconnect reset exponential backoff to zero on each link-dead cycle instead of preserving accumulated retry count. (#5, @v0l)
• FMP/FSP rekey dual-initiation race on high-latency links (Tor): both sides' timers fired simultaneously, both msg1s crossed in flight, each side's responder path destroyed the initiator state. Fixed with deterministic tie-breaker (smaller NodeAddr wins as initiator).
• Parent selection SRTT gate bypass: evaluate_parent used default cost 1.0 for peers filtered out by has_srtt(), defeating the MMP eligibility gate. Now skips unmeasured candidates when any peer has cost data.
• FSP rekey cutover race: initiator cut over before responder received msg3, causing AEAD failures. Fixed by deferring initiator cutover by 2 seconds.
• MMP metric discontinuity after rekey: receiver state carried stale counters across rekey, inflating reorder counts and jitter. Fixed via reset_for_rekey().
• Auto-connect peers exhausted max_retries on initial connection failures and were permanently abandoned. Now retry indefinitely with exponential backoff capped at 300 seconds.
• Control socket permissions: non-root users couldn't connect. Daemon now chowns socket and directory to root:fips group at bind time.
• Post-rekey jitter spikes: old-session frames arriving via the drain window produced 2,000–7,000ms jitter spikes that corrupted the EWMA estimator. Added a 15-second grace period after rekey cutover that suppresses jitter updates until drain-window frames have flushed. (#10)
• ICMPv6 Packet Too Big source was set to the local FIPS address, which Linux ignores (loopback PTB check). Now uses the original packet's destination so the kernel honors the PMTU update. (#16, @v0l)
• Reverse delivery ratio used lifetime cumulative counters instead of per-interval deltas, making ETX unresponsive to recent loss. (#14)
• MMP delta guards used prev_rr > 0 to detect first report, conflating it with a legitimate zero counter. Replaced with has_prev_rr. (#14)

[0.1.0] - 2026-03-12

Added (Initial Release)

Session Layer (FSP)

• End-to-end encrypted datagram service between mesh nodes addressed by Nostr npub
• Noise XK sessions with mutual authentication, replay protection, and forward secrecy
• Automatic session rekeying with configurable time/message thresholds and drain window for in-flight packets
• Port multiplexing for multiple services over a single session
• Session-layer metrics: sender/receiver reports with RTT, jitter, delivery ratio, and burst loss tracking
• Passive RTT measurement via spin bit

IPv6 Adapter

• IPv6 adapter interface allowing tunneling TCP/IPv6 through FIPS mesh for traditional IP applications (TUN interface)
• DNS resolver allowing IP applications to reach nodes by npub.fips name
• Host-to-npub static mappings: resolve hostname.fips via host map populated from peer config aliases and /etc/fips/hosts file

Mesh Layer (FMP)

• Self-organized core mesh routing protocol with adaptive least cost forwarding
• Noise IK hop-by-hop link encryption with mutual authentication and replay protection between peer nodes
• Distributed spanning tree construction with cost-based parent selection and adaptive reconfiguration
• Destination route discovery via bloom filter-based directed search protocol
• Path MTU discovery with per-link MTU tracking and MtuExceeded error signaling
• Link-layer MMP: SRTT, jitter, one-way delay trends, packet loss, and ETX metrics
• Link-layer heartbeat with configurable liveness timeout for dead peer detection
• Epoch-based peer restart detection
• Automatic link rekeying with K-bit epoch coordination and drain window
• Static peer auto-reconnect with exponential backoff
• Multi-address peers with transport priority-based failover
• Msg1 rate limiting for handshake DoS protection

Mesh Peer Transports

• UDP overlay transport with inbound and static outbound peer configuration
• TCP overlay transport with listening port and static outbound peer support
• Ethernet/WiFi transport (MAC address based, no IP stack) with optional automatic peer discovery and auto-connect

Operator Tooling

• Ephemeral or persistent node identity with key file management
• Unix domain control socket for runtime observability
• fipsctl CLI tool for control socket interaction and node management
• Comprehensive node and transport statistics via control socket
• fipstop TUI monitoring tool with real-time session, peer, and transport configuration and metrics display

Packaging and Deployment

• Debian/Ubuntu .deb packaging via cargo-deb
• Systemd service packaging with tarball installer
• OpenWRT package with opkg feed and init script
• Docker sidecar deployment for containerized services
• Build version metadata: git commit hash, dirty flag, and target triple embedded in all binaries via --version

Testing and CI

• Comprehensive unit and integration tests covering all protocol layers and transports
• Docker test harness with static and stochastic topologies
• Chaos testing with simulated severe network conditions: latency, packet loss, reordering, and peer churn
• CI with GitHub Actions: x86_64 and aarch64, integration test matrix, nextest JUnit reporting
• Local CI runner script (testing/ci-local.sh)

Project

• Design documentation suite covering all protocol layers
• CHANGELOG.md following Keep a Changelog format
• Repository mirrored to ngit