Starknet BTC Staking: How to Extract Rewards with Zero Collateral (And Why the Team Ignored It)

The "AI Slop" That Bypassed Your Security
In the world of Web3 security, the "Proof of Concept" (PoC) is the only source of truth. Recently, I discovered a critical logic bypass in the Starknet BTC Staking Attestation module. When I brought it to the team, the response was dismissive: "This looks like AI slop; we won't review it."

Today, I am releasing the technical details and a public PoC to protect the ecosystem. If a vulnerability is "AI slop," then the protocol is currently being exploited by "AI" that knows how to drain its rewards.

The Technical Deep Dive: The "Ghost Staking" Vulnerability
The core of the issue lies in the Attestation.cairo contract. The protocol attempts to verify BTC staking by checking block data, but it makes a fundamental mistake in its "source of truth."

1. The Insecure Root of Trust The contract relies on get_block_hash_syscall(target_block_number) to validate proofs.

Фрагмент кода
// Vulnerable logic snippet
let block_hash = get_block_hash_syscall(target_block_number);
assert(block_hash == provided_proof.block_hash, 'Invalid block proof');
The Flaw: Block hashes are public data. Any network participant can retrieve a valid block hash and include it in a forged attestation. The contract validates that the block exists, but fails to verify that a specific transaction (the BTC deposit) belongs to the user submitting the proof.

1. Lack of Collateral Binding Because the attestation logic doesn't strictly bind a unique BTC transaction to a Starknet address, an attacker can:

Monitor public BTC block data.

Generate a valid-looking attestation structure using public hashes.

Trigger the reward minting function without ever locking a single Satoshi.

Proof of Concept (PoC)
I have isolated the exploit in a clean environment. You can find the full code here: [Insert your GitHub Link: rdin777/starknet-staking_audit1]

To reproduce the vulnerability:

Run Cairo Tests:
cd ghost_staking_audit && snforge test tests/test_ghost_stake.cairo
Result: The test demonstrates a user with zero collateral successfully claiming rewards.

External Simulation:
node scripts/exploit.js
Result: This script generates the forged attestation data that bypasses the assert checks in the contract.

The Disclosure Timeline
March 21, 2026: Submitted the technical report to the team (Lotem Kahana). Dismissed as "AI-generated" without a technical audit.

March 22-24, 2026: Offered private PoC walkthrough. Requested a technical review. Zero response.

March 24, 2026: Public Disclosure.

Conclusion
Dismissing independent research as "AI slop" without running the provided tests is a dangerous precedent. Security is about code, not egos. The vulnerability remains unpatched in the current logic.

I invite the Starknet community to review the PoC and judge for themselves.

Researcher: rdin777
Keywords: #Starknet #Cairo #SecurityAudit #L2 #Vulnerability