CAP Labs — health and slashable collateral disagree

Source & credit. Reproduction of a public audit finding curated by AuditVault — the original finding: 61537-discrepancy-between-health-calculation-and-slashable-collate. Standalone Foundry PoC and full write-up: 61537-health-vs-slashable-collateral_exp in the evm-hack-registry mirror.

Vulnerability classes: vuln/logic/incorrect-state-transition · vuln/logic/state-update · vuln/dos/frozen-funds

Reproduction: Fully local synthetic; run forge test and inspect output.txt. The Playground bundle replays the same 61537-health-vs-slashable-collateral invariant.

Key info#

Field	Value
Loss	Liquidation leaves fresh collateral protected; the protocol absorbs the uncovered amount
Vulnerable contract	Delegation/NetworkMiddleware (reconstructed)
Attacker EOA	0x1111111111111111111111111111111111111111 (synthetic caller)
Attack contract	Exploit
Attack tx	Local `Exploit.run()` call (no historical transaction)
Chain · block · date	Ethereum · local stub 0x1181d03 · 2025-05
Compiler	Solidity 0.8.35 (pragma ^0.8.24)
Bug class	vuln/logic/incorrect-state-transition · vuln/logic/state-update · vuln/dos/frozen-funds

TL;DR#

coverage includes current collateral while slashTimestamp limits slashing to the previous epoch. A fresh rescue deposit improves health but cannot be slashed in liquidation.

Background#

AuditVault finding 61537-health-vs-slashable-collateral is an audit-time issue rather than a historical exploit. This write-up reduces the report to one state transition and keeps the claimed harm assertion executable offline.

The vulnerable code#

The minimized source is explicitly marked RECONSTRUCTED and preserves the report's blamed operation with an @> VULN marker in test/61537-health-vs-slashable-collateral.sol. It is byte-identical to the Playground synthetic source. No verified production source was available in this local checkout.

Root cause#

The two calculations use different time bases: current coverage versus historical slashable deposits.

Preconditions#

The affected protocol path is deployed; the attacker can reach the public operation described in the report. The synthetic removes unrelated integrations while preserving the state and authorization assumptions required for the finding.

Attack walkthrough#

100 mature + 50 fresh collateral gives coverage 150 but only 100 slashable; liquidation leaves the mismatch.
The run() method requires the broken invariant and sets confirmed; the Forge trace records a passing test at output.txt:355.
The remediation is to validate the accounting/state transition before accepting the external call or to consume the one-time state.

Diagrams#

flowchart TD A["Attacker invokes public path"] --> B["Vulnerable state transition"] B --> C["Inconsistent accounting or revert"] C --> D["Reported protocol harm"]

Remediation#

Validate external addresses and returned balances before updating state, and make each one-time transition explicit. Add invariant tests covering zero/deflationary/epoch-boundary inputs.

How to reproduce#

BASH

cd 61537-health-vs-slashable-collateral_exp
forge test -vvv

The browser replay uses scripts/poc-configs/61537-health-vs-slashable-collateral.mjs and the same local-deploy Exploit contract.

Sources#

AuditVault finding
Trail of Bits report
Reduced local source: test/61537-health-vs-slashable-collateral.sol
Forge regression: test/61537-health-vs-slashable-collateral_exp.sol

Sources & further analysis#

Reproductions & code

Standalone PoC + full trace: 61537-health-vs-slashable-collateral_exp (evm-hack-registry mirror).
AuditVault finding: 61537-discrepancy-between-health-calculation-and-slashable-collate.

Alerts & third-party analyses

Web3Sec X hacked database: search.
Rekt leaderboard: search.
Solodit incident search: search.

These dashboards index community alerts tweets, post-mortems, and independent write-ups. Reach them through the protocol name above to cross-check this reproduction against other analyses.

Source-level debugger — step opcodes and Solidity in sync