In a space mission operations context, assets of very high value are operated by several data systems. The security and resilience of these data systems has become a primary concern, due to the increasing ubiquity of the internet and mutability of the threat landscape. Penetration testing is a well established method to identify system security weaknesses, however, is typically high-cost and effort-intensive. The European Space Agency has developed a prototype automated penetration testing framework, called PenBox. Given the volume of parameters and possibilities for the PenBox automated test execution sequence, applied Machine Learning theory presents an interesting opportunity for enhancement. This paper presents the approach and results obtained towards research on the integration of Artificial Intelligence (AI) techniques, in particular Reinforcement Learning (RL), to PenBox. The adopted approach to frame this case study involves the Q-learning paradigm that seeks to learn a policy that maximizes the total reward, according to the quality associated with the certain action to take; the reward itself is the pivotal element that shapes the intelligent agent's behaviour through the trial-and-error theory, without any human interaction. To build this representative space, an existing network simulator was used and further adopted to improve representativeness and alignment with possible PenBox actions. In this paper, we highlight how Deep Q-Learning ensures a certain level of randomization, thus providing the basis for future evolution of the developed AI model to suit the real-world autonomous penetration testing software, with the aim of optimizing the process in terms of performance and cost effectiveness, and enabling an innovative autonomous capability for optimal attack path definitions without human in the loop.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Reinforcing Penetration Testing Using AI


    Beteiligte:


    Erscheinungsdatum :

    05.03.2022


    Format / Umfang :

    1955713 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    REINFORCING UNIT AND REINFORCING MEMBER

    FUNADA RYUICHI / NAM KIWOONG / UENO NORIEDA et al. | Europäisches Patentamt | 2024

    Freier Zugriff

    Modeling and testing of snow penetration

    Lee, Jonah H | Online Contents | 2015


    Reinforcing unit and reinforcing member

    FUNADA RYUICHI / NAM KI-WON / UENO NORIEDA et al. | Europäisches Patentamt | 2024

    Freier Zugriff

    REINFORCING UNIT AND REINFORCING MEMBER

    FUNADA RYUICHI / NAM KIWOONG / UENO NORIEDA et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    REINFORCING UNIT AND REINFORCING MEMBER

    FUNADA RYUICHI / NAM KIWOONG / UENO NORIEDA et al. | Europäisches Patentamt | 2023

    Freier Zugriff