phd placeholder: Accountability Protocol for Future Decentralized Applications

[grimadas]

Working title: Project Noodles

A draft for layers for the accountability tools for the next generation applications. This is semi-layered architecture draft for better understanding of the concept:

• Network layer:
  • Network layer is for building a healthy overlay first and get connected to the giant component of the network;
  • It is used for bootstrap and getting peer list with their latency;
  • Possible strategies: Edge Walk, Random Walk;
  • Using inverted birthday paradox one can also estimate the size of the network(total peers): Peer Counting and Sampling in Overlay Networks:Random Walk Methods
  • Random walks can help avoiding Sybil Regions? (Issue #4481 )
• Overlay layer:
  • Selecting your initial overlay neighbors with some Sybil resiliency guarantees, based on the latency distribution (Telegram is doing that (section 3.3.10));
  • B-Matching could be also used in this layer to optimize peer selection (More details are available here Can's write up);
  • This layer will form a network of semi-stable overlay neighbors that will be used in higher layer.
• Ledger layer:
  • To guarantee tamper resistance we need to store knowledge about other agents and share our world knowledge with others;
  • We need a way to quickly detect double-spends and recover from partitioning while ensuring availability and eventual consistency;
  • The main goal of this layer is to provide quick way for set reconciliation between peers(Efficient way to synchronize and detect inconsistencies);
  • This layer will provide tools for merge semantics, or how to recover from double spends, forks etc.
• Trust layer(or Risk layer):
  • Knowing the overlay and history of transactions from the ledger layer we can have a good estimation on the weighted network topology and information convergence rate;
  • Each transaction from an agent can be assigned with a risk value: e.g. probability of double spending.
• Application layer:
  • We say that security level is application specific, and even agent specific. One can allow risk to lose 1 $, and some only 0.01 $;
  • This layer will define the maximum risk and security level;
  • Related: Risk measure and Value at risk

[grimadas]

Future Post-PhD Roadmap:

This is a first draft planing for my future research/development.

Realizing the dream of 1992 Web of Trust

This involves development, deployment in Tribler and making a research paper on it. This involves:

1. Deep Handshake Integration with IPv8 connection. Based on connection certificates that show the most basic form of trust: connection. A <--> B

2. Trust Crawler: Explore the badges of other nodes to enable indirect trust.

3. Reward Resource Contribution. Trust badges enable the distribution of the donations and rewards and incentives for the contributors.

4. Perfect Meta-data and Tag-Spam. Based on trust we can fight spam and help with the search relevancy.

Non-functional requirements

Infinite interaction until you achieve these:

1. No blocking or annoying Bugs in the code.
2. Don't create resource black hole, keep it light and easy.
3. Clear how to use and support in future.
4. Is able to scale to 1 mln users.