Leakage free helper data storage in microcontroller based PUF implementation

Physical Unclonable Function (PUF) generates a unique identifier of a device, based on variations during a manufacturing process. Such identifier is difficult to predict or clone. It is used usually as a cipher key. Its weakness is, in general, a high sensitivity to changes in environmental conditio...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Microprocessors and microsystems 2021-11, Vol.87, p.103369, Article 103369
Hauptverfasser: Laban, Marek, Drutarovsky, Milos
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Physical Unclonable Function (PUF) generates a unique identifier of a device, based on variations during a manufacturing process. Such identifier is difficult to predict or clone. It is used usually as a cipher key. Its weakness is, in general, a high sensitivity to changes in environmental conditions. In order to ensure stability, post-processing methods with error correction codes are usually applied. Unfortunately, these methods reveal, in many cases, sensitive data. We present a novel way to apply error correction code for the PUF. It is called Code Word Masking construction. This construction allows to generate PUF response in more secure way. Helper data are formed just by properly selecting PUF response bits. Therefore, helper data do not leak sensitive information. The selection is performed according to code words of the error correction code used. The method can be used for any type of weak PUF and many types of error correction codes. The error correction capability of the construction depends only on the capability of the error correction code. We describe this construction, and present an example of the PUF implementation based on the non-initialized values of the static random access memory using a 32-bit microcontroller. The implementation is more secure and has lower entropy loss compared to existing solutions. The reliability of the solution was proved through measurements under various environmental conditions. The implementation is improved by identifying and excluding the unreliable (’dark’) bits.
ISSN:0141-9331
1872-9436
DOI:10.1016/j.micpro.2020.103369