Abstract—Wireless sensor networks (WSNs) are highly sensible to data integrity attacks, which have an important impact on a number of relevant deployments and services. This paper introduces a tolerance approach to fight against data modification attacks in WSNs, which is based on a missing data imputation scheme. The proposal relies on two principal contributions: (1) a multivariate statistical technique where the dynamics of the sensor measurements for the monitored area are captured through the use of dynamic PCA (DPCA), and (2) a variable routing strategy that improves the recovering performance by spreading the effects of the data tampering attack. On the other hand, a complementary multivariate statistical anomaly detection module is implemented to determine the occurrence of data tampering attacks and trigger the subsequent reaction procedure to recover the affected data. As shown by the results obtained, the proposed tolerance approach improves the robustness of a WSN against data tampering attacks, and so its survivability and normal operation over time.

Index Terms—missing data imputation, anomaly detection, Multivariate analysis, response/tolerance schemes, survivability, wireless sensor networks

Cite: Roberto Magán-Carrión, Fernando Pulido-Pulido, José Camacho, and Pedro García-Teodoro, "Tampered Data Recovery in WSNs through Dynamic PCA and Variable Routing Strategies," Journal of Communications, vol. 8, no. 11, pp. 738-750, 2013. doi: 10.12720/jcm.8.11.738-750