Neural networks find applications in many safety-critical systems that raise concerns about their deployment: Are we sure the network will never advise doing anything violating a set of safety constraints? Formal verification has been recently applied to prove whether an existing neural network is certified for some property (i.e., if it satisfies the property for all possible inputs) or not. Formal verification can prove that a network respects the property, but cannot fix a network that does not respect it. In this paper we focus on the automated synthesis of certified neural networks, that is, on how to automatically build a network that is guaranteed to respect some required properties. We exploit a Counter Example Guided Inductive Synthesis (CEGIS) loop that alternates Deep Learning, Formal Verification, and a novel data generation technique that augments the training data to synthesize certified networks in a fully automatic way. An application of a proof-of-concept implementation of the framework shows the feasibility of the approach.

Automated Synthesis of Certified Neural Networks

Zavatteri, Matteo
;
Bresolin, Davide
;
Navarin, Nicolo'
2024

Abstract

Neural networks find applications in many safety-critical systems that raise concerns about their deployment: Are we sure the network will never advise doing anything violating a set of safety constraints? Formal verification has been recently applied to prove whether an existing neural network is certified for some property (i.e., if it satisfies the property for all possible inputs) or not. Formal verification can prove that a network respects the property, but cannot fix a network that does not respect it. In this paper we focus on the automated synthesis of certified neural networks, that is, on how to automatically build a network that is guaranteed to respect some required properties. We exploit a Counter Example Guided Inductive Synthesis (CEGIS) loop that alternates Deep Learning, Formal Verification, and a novel data generation technique that augments the training data to synthesize certified networks in a fully automatic way. An application of a proof-of-concept implementation of the framework shows the feasibility of the approach.
2024
27TH EUROPEAN CONFERENCE ON ARTIFICIAL INTELLIGENCE
27TH EUROPEAN CONFERENCE ON ARTIFICIAL INTELLIGENCE (ECAI 2024)
9781643685489
   iNEST: INTERCONNECTED NORD-EST INNOVATION ECOSYSTEM - Spoke 9 - MODELS, METHODS, COMPUTING TECHNOLOGIES FOR DIGITAL TWIN - AFFILIATO
   iNEST
   Ministero
   ECS00000043

   Future AI Research (FAIR) - Spoke 2 Integrative AI - Symbolic conditioning of Graph Generative Models (SymboliG)
   Ministero
   PE0000013

   Certificazione, monitoraggio, ed interpretabilità in sistemi di intelligenza artificiale
   ISTITUTO NAZIONALE DI ALTA MATEMATICA FRANCESCO SEVERI

   Neuro-Symbolic AI in Digital Twins
   Dipartimento di Matematica “Tullio Levi-Civita”, referente per lo Spoke 9 di iNEST
   40.540,54€
File in questo prodotto:
File Dimensione Formato  
ECAI-2024-Automated-Synthesis-of-Certified-NNs.pdf

accesso aperto

Tipologia: Published (Publisher's Version of Record)
Licenza: Creative commons
Dimensione 304.68 kB
Formato Adobe PDF
304.68 kB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11577/3536903
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? 0
  • OpenAlex ND
social impact