Blockchain technology has created a new cryptocurrency world and attracted a lot of attention. It also attracts scams, for example, phishing scam, a typical fraud, has been found making a notable amount of money in the blockchain ecosystem, which has a very negative impact. Considering the whole life cycle of a phishing scam, this paper proposes the concept of a phishing gang, that is, a set of accounts that serve for phishing activity and belong to the same entity on the blockchain. As phishers often use multiple accounts to commit phishing scams and money laundering, detecting phishing gangs in the blockchain ecosystem is a real and critical problem. To help deal with this issue, this paper proposes a method of detecting phishing gangs on the Ethereum blockchain. Specifically, we first construct a transaction network with a graph structure by mining the transaction record and the account labels of the Ethereum blockchain. Next, we propose the base and improvement methods of taint analysis, aiming to evaluate the taint score of each account by tracking the fund flow of phishing accounts. Then, with the results of taint analysis and some heuristic means, all accounts in the transaction network are divided into five categories. Based on this, we propose a heuristics algorithm for phishing gang detection. And we also summarize gang patterns and reveal money laundering in phishing activities. Experimental results indicate that the proposed framework can be used to build a uniform platform to monitor every account on the Ethereum blockchain for early warning of phishing scams and detection of the phishers' money laundering and cashing process.