EvoSec 2025W Lecture 15: Difference between revisions
Created page with "==Readings== * [https://homeostasis.scs.carleton.ca/~soma/pubs/obada-usec2017.pdf Obada-Obieh, "“Don’t Break My Heart!”: User Security Strategies for Online Dating." (USEC 2017)] * [https://homeostasis.scs.carleton.ca/~soma/pubs/obada-nspw2017.pdf Obada-Obieh, "Can I believe you?: Establishing Trust in Computer Mediated Introductions." (NSPW 2017)] ==Discussion Questions== * How different is dating from other introduction problems today? * How central is reputat..." |
No edit summary |
||
Line 11: | Line 11: | ||
==Notes== | ==Notes== | ||
<pre> | |||
Lecture 15 | |||
---------- | |||
G1 | |||
- dating is long term, but other introductions are more short term | |||
- could star ratings work for dating? probably not...unless looking for hookups | |||
- could penalize new daters who have a bad first date | |||
- reputation is the primary factor in CMIs | |||
- record of many positive interactions -> increased trust | |||
- human->website intro, goes back to narrative auth | |||
- website->website, could use star rep system | |||
G2 | |||
- uber closest to online dating, but time duration is very different | |||
- amount of trust in dating is much higher than every other CMI | |||
- stars definitely influence interactions | |||
- behavior depends on context (kid at school vs at home) | |||
- search engine is a form of CMI | |||
- no matching in the actual ad, but ad targeting is a kind of CMI | |||
G3 | |||
- dating differs in how much info you have to reveal versus other CMI | |||
- reputation is central to CMI - uber, ebay, etc | |||
- ratings are not so objective in dating | |||
- reputation exists all over the place - e.g., online search | |||
- CMI is definitely part of online advertisements | |||
G4 | |||
- dating is a game problem: incentives for honesty, minimal communication, and deception depending on the party | |||
- what are acceptable lies in an online profile? | |||
- applies to other CMI problems | |||
- advertisers are willing to fudge the truth! | |||
- reputation is the most important factor in CMI | |||
- how you calculate trust is very important, flaws will be exploited by adversaries | |||
- search engines & advertisements, both CMIs | |||
Dating: put out info to get a good match, limit info to maintain safety | |||
search engines have different incentives but still there are opposing goals | |||
- web sites don't need to make trust judgements of search engines, different from dating and other CMIs | |||
Question: how could you game dating app reputation systems? | |||
- get your friends to join <- difficult to stop | |||
- sock puppet accounts <- only if auth can be gamed | |||
Why do we care about reputation? When is it important? | |||
- we use it to mitigate risk | |||
(placing ourselves in a vulnerable situation) | |||
- risk & relative power | |||
Reputation is problematic with the Internet today, I think. | |||
- AI | |||
- fraudulent reputation signals (e.g., paid reviews, sock puppets) | |||
- bots, automated attacks on reputation systems | |||
If reputation doesn't scale in an evolutionarily stable way | |||
- efforts to scale reputation will be attacked and circumvented over time | |||
</pre> |
Latest revision as of 19:21, 6 March 2025
Readings
- Obada-Obieh, "“Don’t Break My Heart!”: User Security Strategies for Online Dating." (USEC 2017)
- Obada-Obieh, "Can I believe you?: Establishing Trust in Computer Mediated Introductions." (NSPW 2017)
Discussion Questions
- How different is dating from other introduction problems today?
- How central is reputation to the problem of CMI? To what extent is reputation scalable?
- How generalizable is the concept of computer-mediated introductions? Could a search engine be a form of CMI, but for websites, not people? What about advertisements, such as political advertisements?
Notes
Lecture 15 ---------- G1 - dating is long term, but other introductions are more short term - could star ratings work for dating? probably not...unless looking for hookups - could penalize new daters who have a bad first date - reputation is the primary factor in CMIs - record of many positive interactions -> increased trust - human->website intro, goes back to narrative auth - website->website, could use star rep system G2 - uber closest to online dating, but time duration is very different - amount of trust in dating is much higher than every other CMI - stars definitely influence interactions - behavior depends on context (kid at school vs at home) - search engine is a form of CMI - no matching in the actual ad, but ad targeting is a kind of CMI G3 - dating differs in how much info you have to reveal versus other CMI - reputation is central to CMI - uber, ebay, etc - ratings are not so objective in dating - reputation exists all over the place - e.g., online search - CMI is definitely part of online advertisements G4 - dating is a game problem: incentives for honesty, minimal communication, and deception depending on the party - what are acceptable lies in an online profile? - applies to other CMI problems - advertisers are willing to fudge the truth! - reputation is the most important factor in CMI - how you calculate trust is very important, flaws will be exploited by adversaries - search engines & advertisements, both CMIs Dating: put out info to get a good match, limit info to maintain safety search engines have different incentives but still there are opposing goals - web sites don't need to make trust judgements of search engines, different from dating and other CMIs Question: how could you game dating app reputation systems? - get your friends to join <- difficult to stop - sock puppet accounts <- only if auth can be gamed Why do we care about reputation? When is it important? - we use it to mitigate risk (placing ourselves in a vulnerable situation) - risk & relative power Reputation is problematic with the Internet today, I think. - AI - fraudulent reputation signals (e.g., paid reviews, sock puppets) - bots, automated attacks on reputation systems If reputation doesn't scale in an evolutionarily stable way - efforts to scale reputation will be attacked and circumvented over time