BOOST (Break Our Optical Security Technology) is being organized as the first scientific challenge on optical PUFs (Physical Unclonable Functions). The aim of the challenge is to confirm or disprove the validity of security assumptions that are often made about optical PUF responses. PUFs have emerged as a cost-effective identification/authentication technology, which is urgently needed in the Internet-of-Things (IoT) era. Especially speckle-based optical PUFs, with their unique combination of properties: they are passive, in the sense that the PUF itself contains no electronics and needs no power; they can be read out using a cheap laser and an ordinary smartphone camera; they can be challenged with many different degrees of freedom and contain a large amount of entropy. BOOST is a challenge to the Machine Learning community, aimed to improve our understanding of correlations inside speckle patterns and between speckle patterns.


  • Boris Škorić (TU Eindhoven, The Netherlands)
  • Teddy Furon (IRISA, France)
  • Slava Voloshynovskiy (Univ. Geneva, Switzerland)


    Stage 1:

    Start date: February 1st, 2018
    End date: July 31st, 2018 

    Stage 2:

    Start date: June 1st, 2018
    End date: September 30th, 2018

    Challenge Rules

    Two datasets will be made available

    1. Training data. Query-response pairs from two different PUFs, “A” and “B”. The queries are given as configuration files for a beam-shaping optical element known as Spatial Light Modulator. The responses are given as grayscale images of the response speckle pattern.

    2. Challenge data. A list of queries different from the ones in the training dataset, each query accompanied by a response from either PUF A or PUF B. It is not indicated which PUF.

    The task is to find out, for each query in the challenge dataset, from which PUF the response was obtained. The challenge will have two phases. In the first phase the PUFs are rather thin, which means they contain little entropy. Thus we expect that phase 1 of the challenge is easy. In phase 2 the thickness of the PUFs is increased, resulting in a more difficult challenge. The winner will be the competitor who has submitted the most accurate answers.

  • Objectives and research questions

    Laser speckle has received a lot of attention in the physics community. However, the application of speckle in a security context requires a more multidisciplinary approach. A background in signal processing, computer vision, machine learning and cryptography is indispensable for building scalable authentication/identification protocols and for understanding their security.

    The scientific objectives of the BOOST challenge are:

    1. To get an accurate estimate for the number of independent Challenge-Response Pairs of an optical PUF.
    2. To find out how much Machine Learning can help us in detecting intensity dependencies inside speckle patterns, as well as between different speckle patterns obtained from the same PUF.

    We would like to invite research communities from multiple disciplines to participate in this challenge.