Accountability in Computing
summary
Defines accountability as a trace-driven utility definition and places solutions on a prevention-to-punishment spectrum, formalizing systems as traces plus a violation predicate and utility functions to detect and evaluate punishments. Proves: in open systems with mediated, targeted punishment, either the system must realize principal attribution or it must reduce the violator’s name-level utility enough to offset gains. Compares capture-everything logs, lighter-weight auditing, and shows incentive-compatible protocols can provide automated punishment without identification.
AI-written summary · about
chapters (11)
- 0:00Introduction to accountability
- 2:39Prevention vs. accountability
- 7:21Defining the research scope
- 12:31Formalizing accountability
- 19:15Illustrative car dealership example
- 23:48Review of existing systems
- 27:56Research agenda and challenges
- 32:52Theoretical modeling approach
- 37:37Identity and punishment mappings
- 43:25Open systems and trade-offs
- 49:42Synthesis and future directions
related talks
31mBlame-Preserving Secure Compilation
50mFrom Attribution to Accountability: Upholding International Rules Online
31mDishonest Software: Fighting Back Against the Industry Norms
18mBlame-aware Recomposition for Formally Secure Low-level Compiler Backends
05mConceptualizing attribution and why it matters
36mObject Capabilities
25mImproving Safety with Privacy: An Exploration of Privacy...
48mFormal Modelling of Information-flow Control Systems: Some Lessons Learned