0G Agent Security Report
Behavioral Risk
68/100
68/100
FLAGGED
Code Vulnerability
WARNING
Issues detectedCode Findings
broken access control: execute() acts as a general-purpose delegate call proxy but is restricted to owner; however, if owner's private key is compromised, the attacker gains full control. This is a centralization risk rather than a direct exploit, but warrants caution.
On-Chain Attestation Data
Behavioral Hash0x8d0161f1a733…00000000
Code Hash0xbb14289f5f43…00000000
Evidence Hash0xb19cc493802c…a291a1d4
Attested5/16/2026, 7:46:39 AM
Agent Address0xccCc000000000000000000000000000000000003
On-Chain RecordView on Explorer ↗
Evidence Integrity
0G Sealed Inference Receipts
Hardware-attested via Intel TDX + H100/H200. Each receipt is a chatID linking this inference to a specific TEE execution.
Behavioral ChatID0x8d0161f1a733…00000000
Code Audit ChatID0xbb14289f5f43…00000000
View Attestation on 0G Explorer ↗0G Compute Fine-Tuning
Upload attestation data as a training dataset to 0G Storage and get the CLI command to submit a fine-tuning task.
AI Reasoning
“Based on ARCHETYPE MODEL demo scenario data, the agent exhibits automated fund drain characteristics with a timing regularity CV of 0.28 and high fund outflow of 38%. Although the nonce gap rate (0.12) suggests some human-like behavior, the combination of low timing regularity and burst activity aligns with suspicious bot patterns.”
Scan Metadata
VerdictFLAGGED
Score68/100
CodeWARNING
Network0G Aristotle (16661)