Enigma uses secure addition and multiplication protocols to construct a fully secure interpreter. A user throws a problem to a bunch of decentralized nodes running this secure interpreter that are incentivized by fees to perform these computations and then post the results to a public blockchain, where the validity of the computations can be verified by an auditor.
It also includes (unlike Bitcoin) a distribution protocol so that the "miners" aren't all solving the same problem in a redundant fashion, although I don't think the "network reduction" protocol is specified in the whitepaper.
This is probably the most relevant part to your question:
Code evaluated in our system is guaranteed not to leak any information unless a dishonest majority
colludes (t ≥ n/2). This is true for the inputs, as well as any interim variables computed while the
code is evaluated. An observant reader would notice that as a function is evaluated from inputs to
outputs, the interim results generally become less descriptive and more aggregative.
For simple functions or functions involving very few inputs, this may not hold true, but since these
functions are fast to compute - no additional steps are needed.
However, for computationally expensive functions, involving many lines of code and a large number
of inputs, we can dynamically reduce the number computing nodes as we progress, instead of having
a fixed n for the entire function evaluation process. Specifically, we design a feed-forward network
(Figure 5) that propagates results from inputs to outputs. The original code is reorganized so that we
process addition gates on the inputs first, followed by processing multiplication gates. The interim
results are then secret-shared with N/c nodes, and the process is repeated recursively.
Enigma uses secure addition and multiplication protocols to construct a fully secure interpreter. A user throws a problem to a bunch of decentralized nodes running this secure interpreter that are incentivized by fees to perform these computations and then post the results to a public blockchain, where the validity of the computations can be verified by an auditor.
It also includes (unlike Bitcoin) a distribution protocol so that the "miners" aren't all solving the same problem in a redundant fashion, although I don't think the "network reduction" protocol is specified in the whitepaper.
This is probably the most relevant part to your question:
Code evaluated in our system is guaranteed not to leak any information unless a dishonest majority colludes (t ≥ n/2). This is true for the inputs, as well as any interim variables computed while the code is evaluated. An observant reader would notice that as a function is evaluated from inputs to outputs, the interim results generally become less descriptive and more aggregative.
For simple functions or functions involving very few inputs, this may not hold true, but since these functions are fast to compute - no additional steps are needed.
However, for computationally expensive functions, involving many lines of code and a large number of inputs, we can dynamically reduce the number computing nodes as we progress, instead of having a fixed n for the entire function evaluation process. Specifically, we design a feed-forward network (Figure 5) that propagates results from inputs to outputs. The original code is reorganized so that we process addition gates on the inputs first, followed by processing multiplication gates. The interim results are then secret-shared with N/c nodes, and the process is repeated recursively.