Abstract: A computing machine accesses a set of intermediate artificial neurons in a deep neural network. The deep neural network is fully or partially trained. The computing machine computes, for each artificial neuron in the set of intermediate artificial neurons, an influence score based on an average gradient of an output quantity of interest with respect to the artificial neuron across a plurality of inputs weighted by a probability of each input. The computing machine provides an output associated with the computed influence scores.

Abstract: A security engine may be selected from a plurality of security engines to apply one or more security mechanisms to a section of source code of an application. In some cases, the section of source code may be identified by one or more security mechanism identifiers included in the source code. The security engine may generate machine-readable code that corresponds to the section of source code for which the one or more security mechanisms are to be applied. The machine-readable code may be executed on a plurality of computing devices. In one implementation, applying the security mechanisms to the section of source code may include producing zero-knowledge proofs of knowledge for the section of source code.

Abstract: A security engine may be selected from a plurality of security engines to apply one or more security mechanisms to a section of source code of an application. In some cases, the section of source code may be identified by one or more security mechanism identifiers included in the source code. The security engine may generate machine-readable code that corresponds to the section of source code for which the one or more security mechanisms are to be applied. The machine-readable code may be executed on a plurality of computing devices. In one implementation, applying the security mechanisms to the section of source code may include producing zero-knowledge proofs of knowledge for the section of source code.