Assessing corrosion is crucial in the petrochemical and marine industries. Usual ultrasonic methods based on pulse-echo and guided waves to detect corrosion lack of precision and struggle in structures with a complex shape. In this paper, a complementary and sensitive ultrasonic method based on coda wave interferometry is presented to detect and quantify thickness loss caused by saltwater corrosion of a steel sample. The method consists in exciting the sample and measuring periodically the scattered coda signal. Correlation of the coda signal with a reference taken for the sample initial state permits the monitoring of corrosion spread with a high accuracy. A laboratory experiment is conducted with two steel samples immersed in saltwater with coda and temperature measured simultaneously. One of the samples is protected from corrosion and is used as a control sample to determine the influence of temperature on the coda signals. It is shown that the coda signals on the corroded sample can be temperature-corrected using the temperature measurement only. A control sample is not needed. A good correlation is found between a parameter quantifying the stretching of the coda over time and the corrosion surface, which is monitored with a camera. Finally, a simple theoretical model of coda signal is proposed to quantify the real-time average corrosion rate during the experiment with a sub-micrometric precision. The estimated final average corrosion depth is validated by independent depth profile measurements. The uncertainties and sensitivity of the presented method are investigated.