## Studying Humans in Security How do you study people's use of security? --- ### Contextual Design --- * Go where the user is * Watch what they do * Talk to them --- ### Master and Apprentice model * User is a *master* at being the user * You are an *apprentice*, trying to learn --- ### Principles of Contextual Design * Context * Partnership * Interpretation * Focus --- ### Paper Prototypes * Allows user to do real with with (not yet real) fake system * Allows user to work with real data * Encourages changes to design * Makes users co-designers --- ### Iteration * Go back to the drawing board with new information * Come up with a new prototype * Test (again) with users * Repeat until workable --- ## Challenge **Security** is an *enabling* task, not a production task --- * No one sits down to do "security"* * If you ask someone to, what you get is unrealistic! --- ## How do you study users doing security? --- ### Option 1: Role Play --- * Ask them to pretend to be a fictitious person * Given them some (regular) work for that fictitious person * Introduce attacks, interfaces, or warnings as part of that work --- **Example**: * Pretend to be Bob Jones, the executive assistant of company executive Alice Jackson * Go through Ms. Jackson's email for her. * Decide what needs to be done with each email (response needed, FYI, delete, etc.) * Some email are phishing messages --- **Pros**: * Captures *enabling task* nature of security * Enough control over work to make it someone realistic * Opportunity to introduce new interfaces * Ethical: Not actually attacking users * Easy to set up, run **Cons**: * People can't pretend that well? * Not that realistic; never do work for someone else * Many users quickly figure out deception --- ### Option 2: Embedded Deception --- * Give real employees a new interface to use * *for their real job* * Monitor their work * Without prior warning, do a (fake) attack * See how they react --- **Example** * Ask people to use an upgraded Outlook with a "report phishing" button * Ask to watch them doing their email * Send people a (fake) phishing email * Do they use the button? Why or why not? --- **Pros**: * Very realistic * Get good data about real use * Still can talk to users about it afterwards **Cons**: * Requires deception. Ethical? * Pisses off employees. Breaks bonds with employees. * Data quality? * Not easy to watch users when it happens. (But still have logs) * High effort; can't do many of these --- ## Ethics --- #### Belmont Report * Respect for Persons * Voluntary * Informed Consent * Beneficience * Max benefit, Min harm * Justice * Fair distribution of burdens and benefits --- #### Informed Consent * Subjects should be *informed* of what they are getting into, and still agree to it * Participation should be *voluntary*, and continue to be *voluntary* throughout study * Part of "Respect for Persons": people are autonomous and can make own choices --- #### Deception * A violation of informed consent: subjects don't know something * Only allowable when no other possible way to do it * Also, risks / harms should be balanced with benefits * Requires a *debriefing* and *reaffirmation of consent* * Debrief: Tell subjects they were deceived, full info about study * Reaffirmation: Do they still agree to participate? --- Can you ask subjects to respond to cyber-attacks without deception? --- #### Ethics of Attacking study subjects * Being attacked is scary, sometimes traumatic (harm) * It isn't realistic if you know it is coming (deception) --- #### Ethics of Attacking study subjects * Ensure minimal risk -- no actual damage * Explain as much as possible ahead of time * Debrief and Reaffirm Consent * Provide support, both during and afterwards --- #### Act with Integrity * Handling illegal material? * Conflicts of Interest * Provide data * LLM-generated data? [UXPA Code of Conduct](https://uxpa.org/wp-content/uploads/sites/9/2018/08/CoC_English.pdf)