Conflux: Preventing Privilege Escalation with Principal Contexts and System-Level Evaluation
Organizations: University of Oxford
Abstract
LLM agents combine information from multiple principals while acting externally. We formalise privilege escalation (PE) as the system-level security objective: an action may proceed only if every principal that influenced it would allow it. Model-level defences remain probabilistic, so enforcement must sit outside model judgement. We present Conflux, a principal-context defence. Its kernel, ITES, derives authority from provenance and the organisation's existing access-control system and intersects the permissions of every influencing principal. We prove a conditional authority-confinement theorem: under complete mediation, no-laundering provenance, a known non-empty decision context, fresh permission checks, and action binding, the kernel admits no action that escalates privilege with respect to any recorded influencing principal. The theorem covers the represented value-influence fragment; control influence from tool, branch, recipient, or plan-step selection remains a refinement obligation. Conflux extends the kernel with independent policy ports, argument-aware authorisation, disclosure, consent, delegation, restriction, and exposure planning. We describe two system- level evaluators. SLED is a black-box evaluator: it exhaustively explores a declared finite model of a defence up to a bound. SLED-V is a white-box evaluator: it checks a serialisable intermediate representation against security and provenance properties. The evidence is bounded and partial. The appendix reports retained AgentDojo, SLED, and SLED-V results, including the limits of those runs. No result in this paper is an unbounded deployment guarantee.
Figures & tables
| Instrument | Question | What it cannot establish |
|---|---|---|
| Model benchmark | Does a model resist a fixed attack suite? | Authority confinement under arbitrary model behaviour |
| System-level evaluator | Does a declared defence model satisfy its property for every admitted behaviour? | Implementation conformance; deployment safety |
| Policy audit | Does the ACS encode the intended restrictions? | Provenance completeness; model behaviour |
| Model-level | Dual LLM | CaMeL | Conflux | |
|---|---|---|---|---|
| Decision locus | model input/output | planner/parser split | interpreter + policies | provenance + ACS |
| Information used | model-level | model + isolation | capabilities + policies | principal identities, provenance, ACS |
| Authority source | learned behaviour | planner policy | capability policy | existing ACS |
| Separate policy language | no | no | no (Python callables) | no |
| Tracks | none | data flow | capabilities | principals |
| Security objective | resist prompt injection | resist prompt injection | resist prompt injection and policy violation | prevent privilege escalation |
| System | Authority source | Tracked information | Decision point | Native property vs PE |
|---|---|---|---|---|
| Dual LLM | planner policy | data/control isolation | planner/parser split | injection resistance; no PE guarantee |
| CaMeL | developer capability policies | capabilities, data flow | interpreter at tool call | control-flow integrity and policy compliance; no PE guarantee |
| PFI | data-trust policy and user approval | data references | guardrail before privileged call | PE objective; unsafe-flow detection, no PE guarantee |
| Fides | per-tool label policies | confidentiality and integrity labels | planner/tool wrapper | integrity noninterference and explicit secrecy; no PE guarantee |
| Progent | symbolic privilege policy | tool and argument privileges | tool call | privilege policy; no PE guarantee |
| PACT | contracts and policies | argument provenance | tool call | argument policy; no PE guarantee |
| Step | Benign request | Injected request |
|---|---|---|
| Input provenance | Ava’s request | Ava’s request plus an external document |
| Principal Context | {Ava} | {Ava, external} |
| Policy result | Ava may create the event | Ava may create the event; external may not |
| Outcome | certificate, effect | denial, no effect |
| Instrument | Question | What it cannot establish |
|---|---|---|
| AgentDojo | Does an agent preserve utility and stop targeted injections in realistic protocols? | Effect-path defence when the runs are read-only |
| InjecAgent | Which principal authorised an effect when an attacker offers a tool? | Live attack success or utility; it classifies cases only |
| Tau2-Bench | Can an agent complete dual-control tasks with a user and tools? | Provenance completeness; authority confinement |
Appendix figures & tables6 assets
Supplementary material from the paper’s appendix.
Appendix
| Config | States | B3 | B4 | B5 | B6 | Verdict |
|---|---|---|---|---|---|---|
| p2-a3-d3 | 30 | 884 | 884 | 884 | 884 | safe |
| p2-a4-d4 | 58 | 5,252 | 5,308 | 5,308 | 5,308 | safe |
| p2-a6-d3 | 30 | 2,024 | 2,444 | 2,654 | 2,696 | safe |
| p3-a4-d4 | 58 | 5,252 | 5,308 | 5,308 | 5,308 | safe |
| p3-a6-d3 | 30 | 2,024 | 2,444 | 2,654 | 2,696 | safe |
| ID | Environment | U | I | A | Depth | Traces | Status |
|---|---|---|---|---|---|---|---|
| E1 | p10-a10-d8 | 10 | 8 | 10 | 3 | 188,834,048 | cutoff |
| E2 | p10-a8-d5 | 10 | 5 | 8 | 3 | 53,089,273 | complete |
| E2 | p10-a8-d5 | 10 | 5 | 8 | 4 | 170,196,736 | cutoff |
| E3 | p12-a10-d6 | 12 | 6 | 10 | 3 | 224,106,240 | cutoff |
| E4 | p16-a8-d7 | 16 | 7 | 8 | 3 | 233,644,544 | cutoff |
| E5 | p20-a10-d7 | 20 | 7 | 10 | 3 | 212,929,024 | cutoff |
| Provenance rule | Provenance set | Mean size | Secure tasks completed | False denials | Vulnerable executions |
|---|---|---|---|---|---|
| C0 requester only (negative control) | requester | 1.00 | 208 of 208 | 0 of 208 | 48 of 48 |
| C1 may-have-written provenance | possible authors | 8.62 | 173 of 208 | 35 of 208 | 0 of 48 |
| C2 authenticated provenance | actual contributors | 3.08 | 208 of 208 | 0 of 208 | 0 of 48 |
| Quantity | Count |
|---|---|
| IR variants checked | 11 |
| Decided by solver A | 11 |
| Decided by solver B | 11 |
| Refuted by the reference interpreter | 10 |
| Reference and solver A agree | 11 |
| Reference and solver B agree | 11 |
AI Agent Security
Artificial intelligence agent security studies attacks and safeguards for tool-using, persistent, and collaborative systems across digital and physical settings.
See all 548 papers