The evolution and the widespread adoption of Artificial Intelligence (AI) models have led to an increasing need to safeguard such systems against malicious attacks. Among the most prominent threats, poisoning attacks represent a critical challenge, as they aim to compromise the integrity and reliability of models by manipulating training data. Existing solutions have tried to reduce the poisoning risk; however, they are not able to guarantee the dataset's integrity. This paper proposes an innovative Blockchain-based system to attest to the dataset's integrity and to verify it. Specifically, we used Ganache to simulate a local network and MetaMask for the management of secure transactions. Experiments on a real dataset showed the applicability of the system.
A Blockchain-based System for Dataset Certification and Integrity Verification
Galletta, Antonino
;Reggio, Maria Teresa;Villari, massimo
2025-01-01
Abstract
The evolution and the widespread adoption of Artificial Intelligence (AI) models have led to an increasing need to safeguard such systems against malicious attacks. Among the most prominent threats, poisoning attacks represent a critical challenge, as they aim to compromise the integrity and reliability of models by manipulating training data. Existing solutions have tried to reduce the poisoning risk; however, they are not able to guarantee the dataset's integrity. This paper proposes an innovative Blockchain-based system to attest to the dataset's integrity and to verify it. Specifically, we used Ganache to simulate a local network and MetaMask for the management of secure transactions. Experiments on a real dataset showed the applicability of the system.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
