$DOG
Cooling DownSnapshot Window: 2026-07-25 17:50 UTC · ← Back to Crypto Overview
Social Momentum Summary
Total Engagement - Comments: 0, Retweets: 0, Likes: 0, Impressions: 18
Verbatim Community Citations & Social Evidence 1 source posts analyzed
If you benchmark wallet defenses or phishing detection, you're probably scoring the attacker's tx, not the victim's signature I was measuring whether deployed wallet defenses (tx simulators, address-reputation APIs, rule engines) catch agent-signed drains. Built it on PTXPHISH (NDSS 2025, solid dataset). Then I red-teamed my own setup and the first finding was that I was scoring the wrong transaction. Sharing because it is an easy, invisible mistake. The trap: labeled drainer/phishing datasets record the DRAIN, i.e. the attacker's transferFrom that sweeps the victim's tokens. But a pre-sign wallet defense runs when the VICTIM signs, which is the earlier approve / permit / setApprovalForAll grant. The sweep is a separate tx sent later, by the attacker, from the attacker's address. No wallet defense ever sees it. Feed the sweep to a rule engine and it "catches" the attacker withdrawing to their own address, which is meaningless.