OAK — OnChain Attack Knowledge

Worked example · 2024-07

Gurvinder Bhangu (Gurv) — Celebrity X Account Compromise → Meme Coin Scams — 2024-07

Loss
$530K+ from compromises of Sydney Sweeney and Bob Odenkirk X/Twitter accounts, used to launch Solana meme coin scams. Convicted UK hacker with prior criminal history.
OAK Techniques observed
OAK-T15.006 (Impersonation via Verified Social-Account Compromise) — primary; the verified X accounts of two celebrities were compromised and used to post Solana meme-coin contract addresses to their follower bases. The trust surface is the verification badge attesting authenticity while the attacker holds the account; the followers, not the account owners, are the victims. OAK-T3.003 (Coordinated Pump-and-Dump) — secondary; the operator held the token's pre-allocated initial supply and sold into the post-driven buying pressure within the minutes-long window before recovery. Boundary note: T3.004 is the closer-sounding sibling but is ruled out — its primitive is the promoter's held off-platform reach; this reach was hijacked, not held.
Attribution
confirmed — Gurvinder Bhangu (Gurv), convicted UK hacker. ZachXBT investigation published July 2024.
Key teaching point
The Gurv case demonstrates the celebrity account compromise → meme coin launch pattern: the attacker does not drain the compromised account's followers directly — instead, he uses the celebrity's verified account to post a meme coin contract address. Followers buy the token, believing it's endorsed by the celebrity. The attacker, who pre-minted or pre-allocated the token supply, sells into the celebrity-driven buying pressure. The celebrity's account is the pump mechanism; the meme coin is the extraction vehicle. Detection approach: when a verified celebrity account with no prior crypto content suddenly posts a token contract address, flag it as a likely compromise within seconds — the detection window between the compromise post and the first victim buy is measured in minutes.

Summary

In July 2024, ZachXBT published an investigation into Gurvinder Bhangu (Gurv), a convicted UK hacker, connecting him to the compromises of Sydney Sweeney's and Bob Odenkirk's X/Twitter accounts, which were used to promote Solana meme coin scams.

Attack pattern:

  1. Compromise celebrity X account: Gurv gained access to verified celebrity accounts (Sydney Sweeney — actress; Bob Odenkirk — actor).
  2. Post meme coin contract address: From the compromised account, post a Solana token contract address with a promotional message.
  3. Followers buy the token: The celebrity's millions of followers see the post. Some buy the token, believing it's a legitimate celebrity endorsement.
  4. Attacker sells pre-minted supply: Gurv, who controlled the token's initial supply, sells into the buying pressure, extracting $530K+.

The celebrity account is a "rented" pump mechanism — the attacker uses it for minutes, extracts the value, and abandons it. The account itself is not drained (the celebrity may not even have crypto); it's the FOLLOWERS who lose money buying the scam token.

Timeline

When Event OAK ref
Pre-Jul 2024 Sydney Sweeney and Bob Odenkirk X accounts compromised T6 account takeover
Post-compromise Scam Solana meme coins promoted from celebrity accounts T2 scam token launch
Post-compromise $530K+ extracted from followers buying the scam tokens T2 extraction
2024-07-30 ZachXBT publishes investigation linking Gurv to the compromises (public disclosure)

What defenders observed

  • Celebrity account = distribution channel, not target. The celebrity accounts were not the victims (no crypto was taken from them). They were the DISTRIBUTION INFRASTRUCTURE for the scam — the equivalent of hijacking a TV station to air a fraudulent ad.
  • Meme coin as the extraction vehicle. The scam token is created specifically for this attack. The token has no utility — it exists only to be the asset that victims buy and the attacker dumps. The token contract itself is the scam mechanism.
  • Convicted hacker = identifiable threat actor. Gurv had a prior criminal record, making attribution straightforward once the on-chain trail was connected to known infrastructure.
  • Minutes-long attack window. The celebrity's management/PR team will typically recover the account and delete the scam post within minutes to hours. The attack is designed for this short window — the token surges on the celebrity post, the attacker sells immediately.

What this example tells contributors

  • Celebrity account compromise is a T6 sub-type distinct from exchange account compromise. The attacker is not after the account's crypto assets — they're after the account's DISTRIBUTION POWER (follower count). T6 should distinguish "account takeover for asset theft" from "account takeover as distribution channel."
  • "Celebrity suddenly posts crypto" is a near-zero-false-positive detection rule. When a verified account with no prior crypto content suddenly posts a token contract address, the probability of compromise approaches 100%. This is an automatable detection rule with immediate takedown value.
  • Token contract pre-mint analysis could flag the scam before launch. If the token contract was created minutes before the celebrity post, and the creator wallet holds a large percentage of supply, the scam is detectable before the post goes live. Monitoring newly created tokens where a single wallet holds >50% supply is a pre-attack detection primitive.

Public references

Techniques demonstrated (2)