Artificial Intelligence in the Context of Cybersecurity: Demystifying Machine Learning Through Interactives and Layers of Abstraction

• In Event: Teaching Artificial Intelligence in K–12: Innovative Learning Designs in an Evolving Landscape

Abstract

OBJECTIVE
AI and Cybersecurity are becoming increasingly intertwined, with AI being leveraged for cybersecurity and cybersecurity helping address issues caused by rogue AI (such as deep fakes). The goal in our exploratory project is to dovetail the need to teach these two topics and create a suite of innovative curricular activities [XYZ] (anonymized) that introduce AI/ML in the context of Cybersecurity and prepare highschool teachers to integrate them in their cybersecurity and/or AI curricula. Additionally, [XYZ] activities go beyond the common practice of using extant pre-trained models in an effort to help students build a deeper understanding of how ML works and how the machine actually “learns. We believe that such understanding will aid more meaningful interrogation of critical issues such as ethics and bias in AI models.

THEORETICAL FRAMEWORK & [XYZ] DESIGN FEATURES
[XYZ] is founded on the principle that learning best happens in context (Brown, Collins, & Duguid, 1989), and that when students make real-world connections they’re better engaged and learn more deeply (Engle et al., 2012). In order to make complex ideas of ML (that often involve advanced mathematics) tractable for teen learners, we build intuitions of concepts through games that draw inspiration from Bruner (1960)—“Any subject can be taught effectively in some intellectually honest form to any child at any stage of development.”, and also past work in turtle geometry that made sophisticated mathematics and physics concepts accessible to younger learners through multiple representations and programming (Abelson & diSessa, 1986). In order to help all learners succeed in engaging with and building intuitions about ML algorithms, we scaffold activities using “levels of abstraction” (Waite et al., 2017).

[XYZ] CURRICULAR ACTIVITIES
Our curricular activities introduce core ML topics (such as supervised learning, classification, decision trees, neural networks, optimization, gradient descent, and adversarial examples) contextualized through cybersecurity topics (such as bots, phishing, cyberbullying, deep fakes, cryptography, network security) through a range of unplugged, non-programming and programming activities, and pre-programmed games in an extension of the Snap! block-based programming environment called NetsBlox (Broll et al. 2017).


METHODS & FINDINGS
We conducted a 15 hour pilot online teacher workshop spanning a period of 2 weeks in Fall 2021 with 7 highschool teachers from across the US who taught Cybersecurity and/or AI; 6 of whom completed the pre-post survey and shared feedback (see Fig 1.1, 1.2, 1.3, 1.4). Post-survey responses suggested that teachers found the [XYZ] activities intertwining AI and cybersecurity to be suitable, innovative, and helpful for their own learning. The activities were refined and we have just concluded a week-long teacher summer PD for 6 teachers who will integrate the curricular activities in their cybersecurity curriculum in Fall 2022 and share their experiences from classroom implementation.

SIGNIFICANCE & NEXT STEPS
AI techniques enhance cybersecurity by assisting humans with automated monitoring and responses to cyber attacks. Conversely, cybersecurity can be used to guard against rogue uses of AI. Feedback from our teacher workshops suggest that our approach to meaningfully combine the two topics holds promise.

Authors

• Shuchi Grover, Looking Glass Ventures
  Presenting Author

• Brian Broll, Vanderbilt University
  Non-Presenting Author

• Derek Babb, University of Nebraska - Omaha
  Non-Presenting Author