Skip to content

Latest commit

 

History

History
370 lines (267 loc) · 13.5 KB

File metadata and controls

370 lines (267 loc) · 13.5 KB

Privateness Network - Network Architecture (experimental)

Français

Multi-Layer Protocol Hopping

The privateness.network stack uses protocol hopping across multiple layers. The design aim is to make traffic analysis and tracking harder in practice across all of them at once, but there are no formal anonymity proofs and the real-world effect depends on deployment.

Traffic Flow & Protocol Transitions

Client Application
    ↓ [TCP/IP]
AmneziaWG Access Layer (Obfuscated UDP)
    ↓ [Stealth WireGuard - looks like random data]
Skywire Mesh (MPLS)
    ↓ [Multi-Protocol Label Switching - NOT TCP/IP]
Yggdrasil Overlay (IPv6)
    ↓ [Encrypted IPv6 mesh routing]
I2P Anonymous Network (Garlic Routing)
    ↓ [Layered encryption, multiple hops]
Emercoin Blockchain (Decentralized DNS)
    ↓ [Blockchain-based service discovery]
Destination / Privateness Services

Protocol Layers Explained

Layer 1: AmneziaWG (Access Layer)

Protocol: Obfuscated WireGuard (UDP-based) (see AmneziaWG docs in References / Sources)

  • Obfuscation: Junk packets, header randomization, size variation
  • Appearance: Random-looking data intended to avoid obvious VPN signatures
  • Bypass: Can help with DPI, GFW, and some corporate firewalls (see AmneziaWG documentation for details)
  • Tracking: Aims to make protocol classification and tracking harder in practice

Layer 2: Skywire (MPLS Mesh)

Protocol: Multi-Protocol Label Switching (MPLS)-style mesh (see Skywire docs in References / Sources)

  • NOT TCP/IP: Uses label-switched paths in the mesh core instead of ordinary IP routing
  • Path Selection: Dynamic, multi-path routing
  • Hop Count: Variable, changes per packet
  • Tracking: Removes IP headers in the mesh core, which is intended to make conventional IP-based tracking harder
  • Anonymity: Traffic may be mixed with other users' traffic depending on topology and usage

Layer 3: Yggdrasil (IPv6 Overlay)

Protocol: Encrypted IPv6 mesh (see Yggdrasil docs in References / Sources)

  • Addressing: IPv6 with cryptographic addresses
  • Routing: Distributed hash table (DHT)
  • Encryption: End-to-end encrypted tunnels
  • Tracking: Encrypted mesh with no central routing, intended to make path reconstruction harder compared to a single centralised network

Layer 4: I2P (Garlic Routing)

Protocol: Anonymous overlay network (see I2P docs in References / Sources)

  • Routing: Garlic routing (multiple messages bundled)
  • Tunnels: Unidirectional, frequently rotated
  • Encryption: Layered (like Tor, but with garlic-style bundling)
  • Tracking: Designed to make correlation and tracking harder; if traffic stays fully internal and fully distributed there is no mandatory Tor-style exit-node concept

Layer 5: Emercoin (Blockchain DNS)

Protocol: Blockchain-based naming (see EmerDNS/EmerNVS docs in References / Sources)

  • Resolution: Decentralized, using blockchain records instead of traditional recursive DNS servers
  • Privacy: Avoids traditional DNS leaks for Emercoin-managed namespaces when all lookups go through EmerDNS + dns-reverse-proxy
  • Censorship: Harder to block with conventional DNS/IP blacklists compared to a single centralised resolver
  • Tracking: No single central authority to query for all lookups

How it aims to hinder tracking

1. Protocol Hopping

TCP/IP → Obfuscated UDP → MPLS → IPv6 → Garlic Routing → Blockchain

Each layer uses a different protocol. Tracking requires:

  • Breaking obfuscation (AmneziaWG)
  • Understanding MPLS labels (Skywire)
  • Decrypting IPv6 mesh (Yggdrasil)
  • De-anonymizing garlic routing (I2P)
  • Correlating blockchain queries (Emercoin)

There is no formal bound claimed here on the difficulty of tracking across all layers. The intent is simply to increase the amount of work and visibility an attacker would need across several independent systems.

2. Network Hopping

Entry Node → Mesh Node 1 → Mesh Node 2 → ... → Mesh Node N → Exit
  • Skywire: MPLS path changes dynamically
  • Yggdrasil: IPv6 routing changes per packet
  • I2P: Tunnel hops rotate frequently

Tracking (high-level): May require monitoring many nodes across multiple overlays; this has not been formally analysed.

3. Reduced IP visibility in core

Client IP → [AmneziaWG] → MPLS Labels → [Skywire] → IPv6 Mesh → [Yggdrasil]
  • Skywire core: Uses MPLS-like labels, not IP addresses, in the mesh core
  • No IP headers in core: Traditional IP-layer monitoring is less informative
  • Label switching: Changes at each hop
  • No traceroute: MPLS doesn't respond to ICMP

4. Encryption Layers

[AmneziaWG Encryption]
  └─ [Skywire MPLS Encryption]
      └─ [Yggdrasil IPv6 Encryption]
          └─ [I2P Garlic Encryption]
              └─ [Application TLS/SSL]

5 layers of encryption - breaking one reveals nothing

5. Decentralized Everything

  • No central servers: Can't be subpoenaed
  • No logs: Mesh nodes don't log
  • No DNS servers: Blockchain-based
  • No exit nodes: I2P is fully distributed
  • No ISP visibility: AmneziaWG obfuscation

Attack Resistance

Traffic Analysis Attack

Defense:

  • AmneziaWG obfuscation defeats pattern recognition
  • MPLS label switching breaks IP-based analysis
  • I2P garlic routing mixes traffic
  • Variable packet sizes and timing

Result: Correlating entry/exit traffic should become more difficult in practice, but no anonymity proof is claimed.

Timing Attack

Defense:

  • Multiple protocol layers add variable latency
  • Mesh routing introduces random delays
  • I2P tunnel rotation changes timing patterns
  • Junk packets (AmneziaWG) add noise

Result: Timing correlation attacks may be harder, but this depends on real-world deployment and has not been formally evaluated.

Global Passive Adversary

Defense:

  • MPLS core invisible to IP monitoring
  • IPv6 mesh encrypted end-to-end
  • I2P garlic routing prevents correlation
  • Decentralized architecture (no choke points)

Result: A large-scale adversary would need to observe and analyse several independent layers; this is a design goal rather than a proven property.

Sybil Attack

Defense:

  • Yggdrasil DHT design makes large-scale Sybil attacks significantly harder and more expensive
  • I2P tunnel diversity
  • Emercoin blockchain consensus

Result: Intended to make large-scale Sybil control more difficult and expensive; this does not prove that such attacks are impossible.

Exit Node Monitoring

Defense:

  • I2P has no mandatory exit-node concept like Tor (traffic can remain fully internal; optional outproxies are not required)
  • Yggdrasil mesh is end-to-end encrypted
  • Services hosted on internal Skywire (todo)

Result: When traffic remains internal, there is no single Tor-style exit point; however, other attack surfaces may still exist.

Comparison to Other Networks

vs Tor (informal comparison)

Feature Tor Privateness Network
Entry obfuscation Bridges (detectable) AmneziaWG (stealthy, DPI-resistant in many cases)
Core routing TCP/IP MPLS-style mesh core
Layers 3 (entry, relay, exit) 5+ (AWG, MPLS, IPv6, I2P, blockchain)
Exit nodes Yes (vulnerable) No (I2P internal)
DNS Clearnet DNS (leaks) Blockchain DNS for Emercoin-managed namespaces
Blocking Possible (known IPs) Harder in practice (mesh + obfuscation)

vs VPN (informal comparison)

Feature Commercial VPN Privateness Network
Central servers Yes (single point) No (decentralized mesh)
Logs Often centralized No single place to subpoena (mesh; individual nodes may still log)
DPI detection Easy Harder in practice (obfuscation)
Routing Fixed path Dynamic mesh
DNS VPN DNS (trusted) Blockchain (trustless)
Censorship Blockable Harder to block at scale

vs I2P Alone (informal comparison)

Feature I2P Only Privateness Network
Access layer TCP/IP (detectable) AmneziaWG (stealth)
Mesh routing No Yes (Skywire MPLS)
IPv6 support Limited Native (Yggdrasil)
Blockchain DNS No Yes (Emercoin)
Protocol diversity 1 layer 5+ layers

Real-World Scenarios

Scenario 1: Journalist in Authoritarian Country

Journalist Device
  → AmneziaWG (bypasses GFW, looks like random traffic)
    → Skywire MPLS (no IP routing, untraceable)
      → Yggdrasil IPv6 (encrypted mesh)
        → I2P (anonymous communication)
          → Privateness Services (secure publishing)

Government sees: Random UDP traffic, cannot identify as VPN
ISP sees: Encrypted noise, no protocol signatures
DPI sees: Nothing - obfuscation defeats inspection
Result: Journalist communicates safely

Scenario 2: Corporate Firewall Bypass

Employee Device
  → AmneziaWG (bypasses corporate DPI)
    → Skywire MPLS (exits corporate network)
      → Yggdrasil (encrypted routing)
        → Privateness Network

Corporate firewall sees: Random UDP, not VPN
DPI sees: No VPN signatures
Logging sees: Cannot correlate traffic
Result: Unrestricted access

Scenario 3: Privacy-Conscious User

User Device
  → AmneziaWG (ISP can't see VPN usage)
    → Skywire MPLS (decentralized routing)
      → Yggdrasil IPv6 (mesh anonymity)
        → I2P (garlic routing)
          → Blockchain DNS (no DNS leaks)

ISP sees: Encrypted traffic, no metadata
Advertisers see: Nothing (no tracking possible)
Government sees: Random data
Result: Complete privacy

Technical Deep Dive

MPLS in Skywire

Traditional IP Routing:
Packet → Router 1 (reads IP, routes) → Router 2 (reads IP, routes) → ...
[TRACEABLE: IP headers visible at each hop]

Skywire MPLS:
Packet → Node 1 (reads label, swaps) → Node 2 (reads label, swaps) → ...
[UNTRACEABLE: No IP headers, labels change at each hop]

Label Switching Example

Entry: Label 100 → Node A
Node A: Swap 100 → 200 → Node B
Node B: Swap 200 → 300 → Node C
Node C: Swap 300 → 400 → Exit

IP headers NEVER examined in core mesh

Yggdrasil IPv6 Mesh (Uses a Deprecated IPv6 address range unlikely to be reactivated in the future)

Traditional IPv6: Global routing table, traceable paths
Yggdrasil IPv6: DHT-based routing, encrypted tunnels

Address: 200:1234:5678:abcd::1
  → Derived from public key
  → No geographic information
  → No ISP assignment
  → Fully decentralized

I2P Garlic Routing

Traditional Onion (Tor): Message → Encrypt → Encrypt → Encrypt
Garlic (I2P): Multiple messages bundled, encrypted together

Bundle:
  - Message A (to destination 1)
  - Message B (to destination 2)
  - Dummy message (decoy)
  - All encrypted together

Result: Cannot determine which message is yours

Conclusion

The privateness.network architecture combines several well-known overlay and privacy technologies in a layered way:

  1. Protocol diversity: 5+ different protocols.
  2. MPLS-style core: Label-based forwarding in the Skywire mesh core instead of ordinary IP routing.
  3. Encryption layers: Multiple independent encryption layers.
  4. Decentralization: No single central routing point by design.
  5. Obfuscation: An access layer that aims to reduce the visibility of VPN usage.

The goal is to increase the effort required for large-scale traffic analysis and simple blocking, not to provide mathematically proven untraceability. This should be treated as an experimental architecture whose real-world privacy properties depend heavily on deployment details, correct configuration, and the evolving behaviour of the underlying projects referenced below.

References / Sources