This paper presents a security strategy to protect smart meters against attacks that compromise critical software and hardware components. Many of these attacks consist of replacing or tampering memory components that store measuring procedures and sensitive configuration parameters. Our strategy makes use of physical properties from these components to create a secure identity for the meter. Firstly we explore concepts related to Physical Unclonable Functions to extract unique identifiers from SRAM components. Then we combine these identifiers to create a strong identity. Since this identity depends on intrinsic physical features from the meter components, we can implement a mechanism to detect the attacks described before. We also validate our approach in an experiment using a meter prototype built in an Arduino device and connected to SRAM memories. The results show that our idea is suitable for implementation in smart meters as so as in similar embedded devices.