Psychological Distance in Cyber Decision Making: Information about the Attackers 52nd Edwards Bayesian Research Conference Fullerton, 15 February 2014 Jinshu Cui, Department of Psychology Heather Rosoff, Sol Price School of Public Policy Richard John, Department of Psychology CREATE, University of Southern California Evaluation of Cyber Threats Identity theft? Financial fraud? Computer crash? • Human operators are often thought of as a major cause of security failures - “the weakest link in the chain” [Schneier 2008] • It is difficult for human operators to take cyber threats seriously when few cause serious consequences at the individual level • Critical to understand perception and behavioral response to cyber threats Previous Research • Experience of a near miss significantly increased respondents’ endorsement of safer options, the effect was bigger under a gain frame than a loss frame. • Experience of a hit significantly increased respondents’ endorsement of safer options relative to the near miss past experience. • Experience of a false alarm significantly decreased respondents’ likelihood of endorsing safer response options, compared to the near miss past experience. Rosoff, H., Cui, J., & John, R. S. (2013). Heuristics and biases in cyber security dilemmas. Environment Systems and Decisions, 33(4), 517-529. Real Crime vs. Cyber Crime • Personally targeted • Instant consequences • Have information about the offender, have interaction with the offender, concern about the offender • Group targeted • Delayed consequences • Rarely have information about the attacker, have no interaction with the attacker, ignore the attacker Who? Why? Motivation • Construal level theory (CLT) – “distant” attacks will be viewed abstractly, and “proximal” attacks will be viewed concretely. (Trope & Liberman, 2003, 2010; Trope, Liberman, & Wakslak, 2007) Information about Attackers Attributes Psychological Distance Construal Level attacker unknown most distant highest group distant high individual proximal low physical identified individual most proximal lowest unknown most distant highest terrorism distant high fame proximal low money most proximal lowest identification attacker motivations Experiment 1 – Research Questions • Attacker identification o group or individual o physical identified or not • Attack tactics o personal account o database Experiment 2 – Research Questions • Attacker Motivations o money: purchase luxury items o fame: increase his visibility and reputation within the hacker community o terrorism: provide financial support to a Middle Eastern terrorist group • Resolution Status o resolved o unresolved Experiment 1 - Design • Financial attack scenario • 4 (attacker identification) x 2 (attack tactics) between-subjects design • Manipulations – Attacker identification: • unknown • group • individual • individual with picture – Attack tactics: database vs. personal account Official Bank Notification ___________________________________________________ August 2, 2013 Dear Valued Customer, We are writing to notify you that two days ago, there was an unauthorized attempt to withdraw all of your current funds. (personal account) As of now, we know an individual online hacker is responsible for the breach into your account. (individual attacker) The hacker acted alone in carrying out the attack. We are working with law enforcement officials and regret any concern or inconvenience this incident may have caused you. We will keep you informed as we make progress in his capture. Kindest Regards, Your Bank Experiment 1 – Measures • 10-item PANAS – 1 (not at all) to 5 (extremely) – 5-item negative affect: α = 0.94 – 5-item positive affect: α = 0.84 • 4-item Risk Perception: – 0 to 10 / 0% to 100% – α = 0.83 • 8-item Behavioral Intention: – 1 (strongly disagree) to 5 (strongly agree) – 3-item stay with bank: α = 0.63 – 3-item stay away from bank: α = 0.75 Experiment 1 – Respondents • • • • • • • Recruited from Amazon Mechanical Turk N = 239 $0.55 each Median time to complete: 6 min 43 % female 50% 18-30 years old 98.3% shop online, 92.9% bank online Experiment 1 – Negative Affect Less negative affect associated with pictured individual attacker compared to individual attacker without a picture (p = .038) Mean Negative Affect (1-5) 4.5 Mean Score of Negative Affect attack tactics 4 3.5 database personal 3 2.5 2 individual individual with picture attacker identification low psychological distance would increase participants’ interest in subordinate and secondary aspects (Liviatan, Trope, and Liberman, 2008) Experiment 1 – Positive Affect More positive affect was experienced if a personal account was attacked compared to a database (p = .024) Mean Positive Affect (1-5) 2.8 Mean Score of Positive Affect 2.6 2.4 attack tactics 2.2 2 group 1.8 individual 1.6 individual with picture unknown 1.4 1.2 1 database personal attacker identification Experiment 1 – Protective Behavior When database was attacked, respondents are more willing to count on the bank when the attacker was physically identified; with an individual account attacked, there is little difference. (p = 0.036) Expectation on Bank (1-5) 5 Mean Score of Expectation on Bank 4.9 4.8 4.7 attack tactics 4.6 4.5 4.4 individual 4.3 individual with picture 4.2 4.1 4 database personal attacker identification Experiment 1 – Sex as a Moderator Female respondents tended to experience more negative affect (p = .014), higher perceived risk (p = .022), and were more likely to support for government’s intervention for online protection (p = .021) (Hale, 1996) Experiment 2 - Design • Identity theft scenario • 4 (perpetrator’s motivation) x 3 (resolution status) between-subjects design • Manipulations – Perpetrator’s motivation: • • • • fame money terrorism unknown – Resolution status: • resolved • unresolved • unknown Experiment 2 – Scenes 1 and 2 Scene 1: This morning in the mail you received a credit card statement in your name from a company with which you do not have an account. As you looked over the statement, you noticed several cash advances totaling $500. (PANAS) Scene 2: One week following your receipt of the suspicious credit card statement, you receive the following voice mail: “Good morning, my name is Gabriel Dawson from the Identity Theft Unit of the Police Department. Our investigation into a cyber perpetrator has led us to believe your personal computer has been compromised. We believe this individual hacked into your computer and obtained access to your email account and the cache data of your online activities. In doing so, he was able to obtain your usernames, passwords, banking information, and other personal information. Our investigation thus far shows no evidence that can confirm the perpetrator's intent. (unknown motivation) I plan to be in touch in the coming weeks to report on the progress of our investigation. Please be vigilant in reporting to us any suspicious mail, email, or phone call. Thank you.“ (PANAS, risk perception, short-term behavior) Experiment 2 – Scenes 3 and 4 Scene 3: In the days following the call from the Identity Theft Unit, you notice an increase in suspicious activity. You are receiving more spam emails, junk mails and phone calls from solicitors. More notably is your receipt of a phone call from the Department of Motor Vehicles confirming the issuance of a new driver's license you did not order. You also receive a letter in the mail from the Internal Revenue Service inquiring about your filing of duplicate income tax returns, suggesting that fraudulent returns were submitted in your name. (PANAS) Scene 4: Moving ahead to several weeks following the call from the Identity Theft Unit of the Police Department, you receive yet another credit card statement in the mail from a company with which you do not have an account. This statement has a $1,500 balance. (unresolved) It is clear that you are continuing to experience complications as a result of your identity theft and that you are still at risk. (PANAS, risk perception, long-term behavior) Experiment 2 - Measures • 10-item negative affect (from PANAS): – 1 (not at all) to 5 (extremely) – 8-item negative affect (4 time periods): α = 0.93, 0.92, 0.92, 0.94 • 8-item Risk Perception: – 1 (strongly disagree) to 6 (strongly agree) – 5-item risk perception (2 time periods): α = 0.81, 0.83 • 10-item short-term behavior: – check all that apply – Summed number of checked responses • 12-item long-term behavior: – 1 (strongly disagree) to 6 (strongly agree) – 9-item long-term behavior: α = 0.86 Experiment 2 - Respondents • • • • • • • Recruited from Amazon Mechanical Turk N = 419 $0.75 each Median time to complete: 7 min 44 % Female 50% 18-29 years old 72% have at least one credit card, of which: – 8% have had an account opened fraudulently in their name – 6% pay for an identity theft protection service Experiment 2 – Negative Affect Respondents experienced less negative affect when the identity theft case was resolved compared to unresolved or unknown Negative Affect (1-5) 4.5 Mean Score of Negative Affect (Scene 4) 4 perpetrator's motivation 3.5 3 fame money 2.5 terrorism 2 unknown 1.5 resolved unknown resolution status unresolved Experiment 2 – Risk Perception Respondents perceived less risk of identity theft when the perpetrator was to fund terrorism compared to for money or fame Risk Perception (1-6) Mean Score of Risk Perception 5.2 (scene 2) 5.1 5 4.9 4.8 4.7 4.6 4.5 fame money terrorism resolution status Participants in the low psychological distance condition reported higher risk perceptions (Chandran&Menon, 2004) Experiment 2 – Risk Perception Respondents perceived less risk of identity theft when the situation was resolved compared to unresolved or unknown Risk Perception (1-6) 5.6 5.4 Mean Score of Risk Perception (scene 4) perpetrator's motivation 5.2 money 5 4.8 terrorism 4.6 4.4 4.2 4 resolved unknown resolution status unresolved Experiment 2 – Long Term Protective Behavior Mean Long-term Behavior (1-6) Participants are more willing to pursue long-term behavior of online identity protection when the identity theft case was unresolved or unknown than if it was resolved. Mean Score of Long-term Behavior (Scene 4) 5 4.5 perpetrator's motivation 4 money 3.5 terrorism 3 2.5 2 resolved unknown resolution status unresolved Experiment 2 – Sex as a Moderating Variable Female participants tended to experience more negative affect, high perceived risk, were more likely to seek help (short-term behavior) and more likely to pursue online identity protection (long-term behavior) Conclusions • Cyber attacker and attack characteristics influence respondents’ affective responses, risk perceptions, and intended long term behavior • Cyber Attacker Identification (Individual, Group, Individual with Picture, UK) • Cyber Attack Tactics (Personal account vs. Database) • Cyber Attackers’ Motivations (Fame, Money, Terror, UK) • Resolution of Cyber Attack (Resolved, Unresolved, UK) Psychological Distance in Cyber Decision Making: Information about the Attackers 52nd Edwards Bayesian Research Conference Fullerton, 15 February 2014 Jinshu Cui, Department of Psychology Richard John, Department of Psychology Heather Rosoff, Sol Price School of Public Policy CREATE, University of Southern California Overview • Research Questions – Do attacker identification (e.g., picture or not), attack tactics (i.e., personal account or database), motivations of the perpetrator (e.g., money, terrorism), or resolution of the event influence emotional, cognitive and behavioral responses? • Experiment 1 – Financial Fraud: attacker identification, attack tactics • Experiment 2 – Identity Theft: perpetrator’s motivation, resolution status