A classification framework was developed to identify different types of corrosion in aqueous systems by use of electrochemical noise. Segments of electrochemical noise signals generated by uniform corrosion, pitting corrosion and passivation were analysed in a sliding window. Unthresholded recurrence plots were derived from each segment based on Euclidean distance measures of similarity. Multivariate image analysis was used to extract features from the recurrence plots and these features were used as predictors in a classifier trained to identify the corrosion types. The classification models showed satisfactory prediction accuracy and could potentially be applied in real-time corrosion monitoring systems. The length of the sliding window is a critical parameter in the proposed framework and its effect on classification performance was investigated as well. © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.