On the Effect of Aging on Digital Sensors

Sensing environmental variables is an important step for autonomous devices which operate in different conditions. For this purpose, the digital chips embed sensors, preferably digital for portability reasons, to ensure that they are not operating out-of-specifications. However, a digital sensor is exposed to various stress as well as aging during its lifetime. Aging makes the sensors’ output deviate from the original intended one, resulting in an incorrect report of the operating condition and in turn improper actions based on such reports. This paper presents a thorough study on the effect of aging on digital sensors. We first propose two low-overhead characterization schemes for these sensors. Then, we investigate the accuracy of such characterizations on different processes, voltages, temperatures and aging conditions. Our results show that sensors’ ability to detect attacks (as well as false detections) varies with aging, notably in the first 5 years of usage. Indeed, comparing new versus aged sensor shows that 3% to 12% of the alarms that would raise if the sensor were new, will not be fired for a 5 year old device, while 15%-25% false alarms are raised for the device after 5 years of usage. This brings a great concern to the reliability as well as the security of the digital sensors which can not be used as is after aging. It is also shown that these results stay the same whatever the variation of the process. Consequently, it is recommended to associate an aging correction to digital sensors, such as in-field calibration techniques, and/or sensor replication.