A carregar...
Publicação
Formalisation of Smart Contract Languages
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.38 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Smart contracts automatically verify and enforce contractual agreements without the
need of a trusted intermediary, as potential conflicts are resolved by the network’s consensus
protocol. Since "code is law", contracts should be correct, but bugs and vulnerabilities,
often exploited by attackers, allow erroneous or even fraudulent behaviour. These days
smart contracts are still mostly being written in general purpose programming languages,
without proper specifications, let alone correctness proofs. Immutability is one of its
selling points, but it is also one of its major problems, as once a contract is deployed to
the blockchain it cannot be amended. Additionally, many vulnerabilities come from the
misimplementation of contracts’ intended behaviour, as developers struggle to grasp the
behavioural impact that the contract has in the blockchain. It is thus crucial to achieve
correct implementations of smart contracts.
In order to aid developers to promote the design of safer contracts that follow the
protocols they are supposed to implement, we propose the use of behavioural types in
smart languages. We believe that the use of typestates, for dynamic checking, and session
types, for static checking, can ensure the intended behaviour of the contract before and
during its execution.
To better understand a contract’s behaviour throughout its execution, we took advantage
of Racket (and PLT Redex), to have a visualisation of a step-by-step execution graph.
By formally defining the syntax and reduction rules of a "core" smart contract language,
and how each rule affects the statements and the programs configuration, this visualisation
tool allows programmers to check and adjust the language’s formal semantics. This is
a successful proof-of-concept exercise, confirming the suitability of Racket to develop program
semantics which can be analysed throughout its execution. In the context of smart
contract languages, these are important features (to be combined with formal verification
with proof assistants). Furthermore, we also implemented a typechecker in OCaml that
provides a type derivation tree of the program, in addition to preventing the occurrence
of execution errors.
To illustrate the usefulness of this approach, we took two different smart contract
languages, one completely formalised with syntax, operational semantics and type system (Featherweight solidity, FS), and another only with its natural language semantics (Flint).
We formalised FS in Racket and OCaml, where we were able to detect an inaccuracy;
and we repeated this process with Flint, formalising its operation semantics and type
system. The latter was much more challenging as it incorporated the use of typestates.
Throughout this thesis, we present many examples on how the use of visual tools can
help in the developing states of contracts and better understand the correct execution of
programs, as well as how the use of behavioural types can prevent many execution errors
even before running.
The framework we define herein not only finds defects in the contracts, but also,
crucially, detects vulnerabilities in a language construction, as we demonstrate with our
use-cases. Therefore, this approach is very valuable not only for the programmer as visual
debugging, but also for the language designer to test the effects of definitions.
Descrição
Palavras-chave
Executable Operational Semantics Error Detection Smart Contract Languages Behavioural types Type Systems Programming Language Formalisation