3 of 59

Why do we need it?

Consensus in a trustless, decentralized network

Decentralized services

Remote computation

And more...

4 of 59

In the context of Golem

5 of 59

In the context of Golem

6 of 59

In the context of Golem

7 of 59

What is Golem?

A network of heterogeneous resources that can be either used by a requestor or provided to other participants by a provider.

Additional layers on top of the infrastructure (e.g., economy).

8 of 59

Problem statement

Requestor wants to be sure that

He receives valid results from the provider after a trustworthy execution on the provider’s machine

Gets results within a reasonable time

9 of 59

Problem statement

Provider on the other hand

Needs to be protected from malicious tasks

10 of 59

High-level requirements

Requestor should be able to run an arbitrary, unmodified binary efficiently

11 of 59

High-level requirements

Requestor should be able to run an arbitrary, unmodified binary efficiently

Task input data has to be the same as prepared by the requestor

12 of 59

High-level requirements

Requestor should be able to run an arbitrary, unmodified binary efficiently

Task input data has to be the same as prepared by the requestor

The binary run by a provider must be the one requestor wanted

13 of 59

High-level requirements

Requestor should be able to run an arbitrary, unmodified binary efficiently

Task input data has to be the same as prepared by the requestor

The binary run by a provider must be the one requestor wanted

The execution of the task has to be carried out without tampering

14 of 59

High-level requirements

Requestor should be able to run an arbitrary, unmodified binary efficiently

Task input data has to be the same as prepared by the requestor

The binary run by a provider must be the one requestor wanted

The execution of the task has to be carried out without tampering

Output data cannot be altered without detection

15 of 59

High-level requirements

Requestor should be able to run an arbitrary, unmodified binary efficiently

Task input data has to be the same as prepared by the requestor

The binary run by a provider must be the one requestor wanted

The execution of the task has to be carried out without tampering

Output data cannot be altered without detection

Only the requestor can see the plain text input and output data�

16 of 59

High-level requirements

Requestor should be able to run an arbitrary, unmodified binary efficiently

Task input data has to be the same as prepared by the requestor

The binary run by a provider must be the one requestor wanted

The execution of the task has to be carried out without tampering

Output data cannot be altered without detection

Only the requestor can see the plain text input and output data�

Provider has to be protected from malicious binaries

17 of 59

Meeting the requirements

Limiting the class of problems

deterministic (PoW or verification)

nondeterministic (probabilistic verification)�

By means of the infrastructure (TEEs)

Introducing external sources of trust

Hybrid approach

18 of 59

Choosing the best approach

Tradeoff

Inherent to the method

Task specific

Runtime dependent

Focus on a single one

Based on the hardware infrastructure

Allows remote computations

19 of 59

Meeting the requirements

SGX, Graphene and Graphene-ng

20 of 59

SGX - short recap

Overview

An Intel® architecture extension designed to increase the security of application code and data

Enclave based computation model (application is run in containers called enclaves, which are supposed to be protected from the host)

Provides a mechanism for remote attestation

Developer perspective - an enclave

Static interface specification

Source code + Intel SDK

By default limited to computations

Host (Untrusted Environment)

21 of 59

SGX - short recap

Summary

Provides important security features

Does not allow to run arbitrary applications in the enclave

May be treated as a building block for more generic solutions

Host (Untrusted Environment)

22 of 59

Graphene overview

Features

A framework with LibOS, which simulates the OS for the app in the enclave

Runs unmodified, full Linux apps protected from the host

Makes use of Intel SGX�

Graphene architecture

23 of 59

Graphene overview

Graphene vs other approaches

TCB vs attack surface �
Flexibility - other host or guest OS can be supported by changing a single component

Better suited to sandboxing�

Graphene architecture

24 of 59

Graphene overview

With Graphene, a developer doesn’t have to

Rewrite the code to suit SGX needs

Specify the static enclave interface

Compile the enclave

Default SGX enclave building process

25 of 59

Graphene overview

With Graphene:

An arbitrary, unmodified Linux binary can be run inside an enclave

26 of 59

Graphene overview

With Graphene:

An arbitrary, unmodified Linux binary can be run inside an enclave

The process still requires manual work to configure, deploy and run the application

27 of 59

The next step

Graphene-ng

28 of 59

The next step

Graphene-ng = Graphene

+ Protected files

�+ Docker support

+ Tools (scripts)

+ Bug fixes�

29 of 59

The next step

Graphene-ng = Graphene

+ Protected files

�+ Docker support

+ Tools (scripts)

+ Bug fixes�

Resulting in a better UX regarding interaction with the enclave for both, the developer and the end user.

30 of 59

Graphene-ng by example

PoC integration with Golem

31 of 59

Provider setup

SGX must be enabled on provider's machine��Protected Files configuration:

Enclave manifest

Running the docker image

Graphene-ng Docker configuration

32 of 59

Provider setup

Graphene-ng Docker for Brass Golem configuration

SGX must be enabled on provider's machine��Protected Files configuration:

Enclave manifest

Running the docker image

33 of 59

Initialization - setup and handshake

The process

Semi-automatic

Uses convenience scripts

34 of 59

Application POV

35 of 59

Requestor POV

Requestor looks for SGX nodes in a P2P network

36 of 59

Requestor POV

And connects to available nodes that offer the SGX architecture

37 of 59

Requestor POV

It can be envisioned as if the local machine had more compute resources available locally

38 of 59

And yes, it works

Benefits for Golem and the community�

Confidential remote computation

Using unmodified binaries

SGX as an infrastructure in Golem�

39 of 59

Next steps

SGX:

Liberation: IAS, FLC

Mitigation of known attacks

EPC size and efficiency�

Development:

Stable and efficient Graphene-ng

Proofs of computation

Windows support�

40 of 59

Graphene-ng Use Cases

Golem ecosystem

41 of 59

Local Verification

Task is computed by an untrusted host

And verified in an enclave run by the provider

42 of 59

Golem Unlimited

Infrastructure based on a LAN-like network setup

Two trusted actors:

The hub - requestor role

Provider nodes

43 of 59

Golem Unlimited - powerful component with SGX

Each Unlimited provider exposes SGX backend

Can be envisioned as a powerful component with SGX added to the Golem network

44 of 59

Golem Unlimited - storing identities

Storing the identity information only inside enclaves

Binds identities to physical machines in a passwordless manner

45 of 59

Use Cases - other Golem integrations

Integrations with Golem by external developers

46 of 59

Graphene-ng Use Cases

02. Wider audience

47 of 59

Decentralized server implementation

Centralized infrastructure

48 of 59

Decentralized server implementation

Decentralized architecture

May be potentially used to facilitate secure multi-party computation

49 of 59

Atomic swap

Cross-blockchain atomic swap

Logic implemented inside Enclave

50 of 59

Distributed exchanges

Implemented by means of cross-blockchain atomic swap

51 of 59

Storing secrets

Sealing private keys locally

Hardware wallet primitive

Sealing private keys remotely

52 of 59

Storing secrets

Threshold signatures

Identity stored in a decentralized way

53 of 59

Graphene-ng Use Cases

03. In existing projects

54 of 59

Minimal Viable Plasma

Minimal Viable Plasma with operator

55 of 59

Minimal Viable Plasma

Plasma chain operator can be implemented in an SGX enclave

56 of 59

Minimal Viable Plasma

Plasma chain operator can be implemented in an SGX enclave

And additionally decentralised

58 of 59

Data streaming

59 of 59

Thank you

GOLEM.NETWORK

PIOTR JANIUK

contact@golem.network

1 of 59

2 of 59

3 of 59

4 of 59

5 of 59

6 of 59

7 of 59

8 of 59

9 of 59

10 of 59

11 of 59

12 of 59

13 of 59

14 of 59

15 of 59

16 of 59

17 of 59

18 of 59

19 of 59

20 of 59

21 of 59

22 of 59

23 of 59

24 of 59

25 of 59

26 of 59

27 of 59

28 of 59

29 of 59

30 of 59

31 of 59

32 of 59

33 of 59

34 of 59

35 of 59

36 of 59

37 of 59

38 of 59

39 of 59

40 of 59

41 of 59

42 of 59

43 of 59

44 of 59

45 of 59

46 of 59

47 of 59

48 of 59

49 of 59

50 of 59

51 of 59

52 of 59

53 of 59

54 of 59

55 of 59

56 of 59

57 of 59

58 of 59

59 of 59