In the landscape of cybersecurity, a large field of research is dedicated to physical attacks since the publication of the first attacks in the early 1990s. Side-channel attacks can reveal the secret values processed in a circuit by observing physical quantities (power consumption, electromagnetic emissions, execution time, etc.). Physical attacks constitute an important threat against embedded systems; in particular, they are the most effective way to break implementations of cryptography. The Smart Cards industry is up with the design of countermeasures, and high security products embed a large set of hardware and software countermeasures. With the emergence of the Internet of Things, we observe a rapid increase of the number of communicating devices, which present various security needs, but also unequal levels of security [7]. Hence, we advocate for the design of tools to automate the application of Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. In the COGITO project, we focused on the use of runtime code generation to introduce behavioural variability in embedded systems. Indeed, behavioural variability is often used as a protection against physical attacks [6]. Security products embed hardware and software desynchronisation mechanisms to achieve variability in side-channels: for example clock jitters in hardware or dummy loops of random duration in software. We defined code polymor-phism as the capacity of a program component to vary its observable behaviour, at runtime, without altering its functional properties. Code polymorphism can be considered as a hiding countermeasure: the information leakage, which is observable physical quantities during the secured computation, is hidden in the information noise produced by the behavioural variability generated by the poly-morphism. However, code polymorphism alone does not remove information leakage as it would be the case with masking countermeasures. We implemented code polymorphism with runtime code generation of machine binary instructions (Fig. 1): the polymorphic component is composed of (1) dedicated runtime code generators, specialised for the targeted component so that it presents a low memory footprint and a short code generation time, and (2) of polymorphic instances which are the many code variants produced by the polymorphic code generator at runtime. In order to produce many code variants of the same functional component, the runtime code generator is driven by a source of random data. The successive execution of many polymorphic instances, which are all functionally equivalent but composed of different series of machine instructions, will induce a strong variability in the observable behaviour of the polymorphic component.