Sensitive Samples Revisited: Detecting Neural Network Attacks Using Constraint Solvers

TL;DR

This paper presents a novel method for detecting neural network attacks by modeling sensitive samples with symbolic constraint solvers, offering a flexible and complete alternative to gradient-based approaches.

Contribution

It introduces a constraint solver-based approach for identifying sensitive samples, improving flexibility and completeness over previous gradient ascent methods.

Findings

Effective detection of Trojan attacks demonstrated

Constraint solver approach outperforms gradient-based methods in flexibility

Partitioning search space improves efficiency and completeness

Abstract

Neural Networks are used today in numerous security- and safety-relevant domains and are, as such, a popular target of attacks that subvert their classification capabilities, by manipulating the network parameters. Prior work has introduced sensitive samples -- inputs highly sensitive to parameter changes -- to detect such manipulations, and proposed a gradient ascent-based approach to compute them. In this paper we offer an alternative, using symbolic constraint solvers. We model the network and a formal specification of a sensitive sample in the language of the solver and ask for a solution. This approach supports a rich class of queries, corresponding, for instance, to the presence of certain types of attacks. Unlike earlier techniques, our approach does not depend on convex search domains, or on the suitability of a starting point for the search. We address the performance limitations of constraint solvers by partitioning the search space for the solver, and exploring the partitions according to a balanced schedule that still retains completeness of the search. We demonstrate the impact of the use of solvers in terms of functionality and search efficiency, using a case study for the detection of Trojan attacks on Neural Networks.