____
| _ \ __ ___ _____
| | | |/ _` \ \ / / _ \
| |_| | (_| |\ V / __/
|____/ \__,_| \_/ \___|
Anonymised peer-to-peer distributed hash table on proof-of-work and gossip.
Copyright 2024 Joey Innes <joey@inneslabs.uk>
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Dave is an open protocol, designed to simply & efficiently distribute a hash table in a byzantine environment.
A peer-to-peer cache, anonymised, and almost instantly-consistent.
Clients (webapps included, via HTTP gateway), may write to the network without any token or key, providing only a classic hashcash style proof-of-work.
Packets with fields value, time, and a time-bound proof-of-work (nonce, work) are gossiped, until their weight has decayed, such that they are no-longer available on any node.
1. Constants
MTU 1500 Reasonable value to avoid packet fragmentation.
NPEER 2 Limit to prevent Eclipse attack.
EPOCH 26544358ns Base minimum period of the operation cycle.
DELAY 512 Epochs until dats are shared to new peers.
SHARE 8 Epochs until silent peers are no-longer advertised.
PING 64 Epochs until silent peers are pinged.
DROP 512 Epochs until silent peers are dropped from the peer table.
PRUNE 128 Number of epochs between dat pruning.
SEEDSEED 1024 For seed nodes. Number of epochs between sending a random dat. Seed nodes must already send many PEER messages, but also help to seed.
2. User-configurable Settings
SUGGESTED VALUE
FILTER_CAP 1M Allocates around 1MB. Should be increased to 2, 4 or 8M if possible.
DAT_CAP 1M As large as possible for available memory. ~1304B per dat.
3. Operation Codes
GETPEER Packet requesting that the remote replies with NPEER random peer descriptors.
PEER Packet containing NPEER peer descriptors.
DAT Packet containing a value, time, and output of the cost function: work, salt.
4. Types
4. 1. M
dave.M is the binary serialisation of an entire packet.
DESCRIPTION BYTE LENGTH
Op Operation code. 1
Peers List of peers. 2*NPEER
Work SHA256(SHA256(Val, Time), Salt). 0 | 32
Salt Random bytes used to solve work. 0 | 32
Time Big-endian unix milliseconds. 0 | 8
Val The data. 0 | <= 1388 when NPEER=2
4. 2. Dat
godave.Dat is a container for Value, Time, Salt, and Work.
The proof of work allows the network to prioritise storage of newer keys backed by more work. See 6.
4. 3. Wire Format
A message is serialized into binary using protobuf. See protobuf spec dave.proto.
Transpiling Protobuf Spec for Go:
#!/bin/bash
protoc --go_out=. dave.proto
5. Peer Discovery & Liveness
The protocol ensures a cohesive network by combining liveness and peer discovery into a single pair of direct messages (GETPEER & PEER).
A node replies to a GETPEER message with a PEER message with up to NPEER peer descriptors.
Peers are not advertised if they have not been seen recently.
6. Mass
mass = difficulty * (1 / ageMilliseconds)
Where difficulty is number of leading zero bytes, age is calculated from the given time and current time.
7. Replacement by Mass
Once per PRUNE epochs, a user defined number of heaviest dats are kept, and the remaining dropped.
8. Send One Random Dat
Every EPOCH, each node sends one random dat to one random peer. This ensures reliable distribution and sender anonymitiy.
Propagating dats in this way ensures that an adversary is unable to create a timing-attack to discern the source of a dat, even with a broad view of network traffic.
9. Packet Filter
Dropping packets efficiently is fundamental to resilience against DoS attack.
Cuckoo filters leverage cuckoo hashing to efficiently store fingerprints in a compact hash table, enabling fast insertions, deletions and lookups.
This makes them well-suited for performance-critical applications like packet filtering to improve DoS attack resilience.
Packets that deviate from the protocol are detected & dropped without further processing.
The key inserted uniquely into the cuckoo filter:
FNV128A(REMOTE_IP, HASH4(PORT), OP_CODE)
Failing unique insertion, the packet is dropped.
The cuckoo filter is reset every epoch, therefore each OP_CODE may be sent once per IP-PORT per epoch.
The number of ports allowed per IP address is limited using a 4-bit multiply-then-shift hash function.