Ethereum-Analysis

Analysis of Etherum contracts, transactions, gas, scams and scammers’ graph using spark and graphframes

• Total number of transactions

• Top 10 most popular services

• Most lucrative forms of scam

• How different scams changed with time

• Transactions gas price vs time

• Contract gas vs time

• Contract complexity vs time

• Triangle count

• Scammer wallets used for accumulating Ether

Dataset

The dataset was collected using the dumps uploaded to a repository on BigQuery. A subset of this data , including contracts and transactions, was then uploaded to HDFS at /data/ethereum. Also, a set of identified scams run on the Ethereum network were collected using etherscamDB and uploaded to HDFS at /data/ethereum/scams.json.

Dataset Schema

• Blocks

  +-------+--------------------+--------------------+----------------+-----+---------+--------+----------+-----------------+
  | number|                hash|               miner|      difficulty| size|gas_limit|gas_used| timestamp|transaction_count|
  +-------+--------------------+--------------------+----------------+-----+---------+--------+----------+-----------------+
  |4776199|0x9172600443ac88e...|0x5a0b54d5dc17e0a...|1765656009004680| 9773|  7995996| 2042230|1513937536|               62|
  |4776200|0x1fb1d4a2f5d2a61...|0xea674fdde714fd9...|1765656009037448|15532|  8000029| 4385719|1513937547|              101|
  |4776201|0xe633b6dca01d085...|0x829bd824b016326...|1765656009070216|14033|  8000000| 7992282|1513937564|               99|
  

• Transactions

  +------------+--------------------+--------------------+-------------------+------+-----------+---------------+
  |block_number|        from_address|          to_address|              value|   gas|  gas_price|block_timestamp|
  +------------+--------------------+--------------------+-------------------+------+-----------+---------------+
  |     6638809|0x0b6081d38878616...|0x412270b1f0f3884...| 240648550000000000| 21000| 5000000000|     1541290680|
  |     6638809|0xb43febf2e6c49f3...|0x9eec65e5b998db6...|                  0| 60000| 5000000000|     1541290680|
  |     6638809|0x564860b05cab055...|0x73850f079ceaba2...|                  0|200200| 5000000000|     1541290680|
  

• Contracts

  +--------------------+--------+---------+------------+--------------------+
  |             address|is_erc20|is_erc721|block_number|     block_timestamp|
  +--------------------+--------+---------+------------+--------------------+
  |0x9a78bba29a2633b...|   false|    false|     8623545|2019-09-26 08:50:...|
  |0x85aa7fbc06e3f95...|   false|    false|     8621323|2019-09-26 00:29:...|
  |0xc3649f1e59705f2...|   false|    false|     8621325|2019-09-26 00:29:...|
  

• Scams

  0x11c058c3efbf53939fb6872b09a2b5cf2410a1e2c3f3c867664e43a626d878c0: {
    id: 81,
    name: "myetherwallet.us",
    url: "http://myetherwallet.us",
    coin: "ETH",
    category: "Phishing",
    subcategory: "MyEtherWallet",
    description: "did not 404.,MEW Deployed",
    addresses: [
        "0x11c058c3efbf53939fb6872b09a2b5cf2410a1e2c3f3c867664e43a626d878c0",
        "0x2dfe2e0522cc1f050edcc7a05213bb55bbb36884ec9468fc39eccc013c65b5e4",
        "0x1c6e3348a7ea72ffe6a384e51bd1f36ac1bcb4264f461889a318a3bb2251bf19"
    ],
    reporter: "MyCrypto",
    ip: "198.54.117.200",
    nameservers: [
        "dns102.registrar-servers.com",
        "dns101.registrar-servers.com"
    ],
    status: "Offline"
  },
  

  Analysis & Results

1. Total number of transactions

In number_of_transactions.pythe total number of transactions are aggregated for each month included in the dataset.

Results:

2. Top 10 most popular services

In top10_addresses.py the top 10 services with the highest amounts of Ethereum received for smart contracts are yielded.

Results:

3. Scams

3.1. Most lucritive forms of scams

In scams.py, the types of scam making the most Ether are yielded.

Results:

3.2. How different scams changed with time

In scams.py the total sum of Ether made off of each scam category is calculated for each month included in the dataset.

Results:

• In July of 2017, phishing attacks were the most profitable form of scam.
• In September of 2017, fake ICO was the most profitable form of scam.
• In September of 2018, scamming attacks were the most profitable form of scam.

4. Gas

4.1. Transactions gas price vs time

In gas.py, the average price of gas is calculated for each month by: total sum of gas prices / total number of transactions.

Results:

• The average price of gas has generally decreased with peaks during the first few months of each year.

4.2. Contract gas vs time

In gas.py, the same approach for transactions is taken for smart contracts as well, with one alteration: in order to make sure a block represents a smart contract, it needs to be joined with the contracts dataset. As the block id is unique and present in both datasets, it is used as the joining key.

Results:

• Contracts have been requiring more gas since the start of Ethereum, and it appears that the needed gas has reached a somewhat steady state.

4.3. Contract complexity vs time

The complexity of a transaction is reflected in its difficulty level, which is available in the blocks dataset.

Results:

• Contracts have become more complicated with peaks in late 2017 and 2018, followed by a general decrease.

• The required gas for a contract apears to be strongly correlated with the contract’s complexity (difficulty):

corr(diff, gas) = 0.9385

• With higher complexity, more gas would be required, so the most popular services might have been mining complex contracts.

5. Graph Analysis

5.1. Triangle count

The target dataset for this part of the analysis is the transaction dataset. To find all the nodes in the dataset (to addresses and from addresses), the to and from addresses are concatenated using the union() method. Then, to avoid repetition of nodes, the distinct() method is called on the vertices RDD. Next, the RDDs are converted to dataframes (with the correct formats described in e_f() and v_f()) and the graph is built using these two dataframes. Finally to find the triangles, a motif is called on the graph, searching for all sets of 3 nodes that form a triangle: a–>b, b–>c, c–>a.

Results:

The resulting file is too large to include here.

5.2. Scammer wallets

In scammers_graph.py, the general premise is that the addresses where scammers are accumulating their stolen cryptocurrency, has a lot more inDegrees than outDegrees, and that there are fairly many inDegrees. These assumptions are used in the custom function is_scammer(). The inDegrees and outDegrees are calculated for all of the nodes and converted into RDDs to allow RDD transformations. To calculate the ratio of outDegrees to inDegrees for each address, the inDegrees and outDegrees RDDs are joined with the address as the join key. Then, the joined RDD is filtered using is_scammer() to see which addresses meet the conditions stated above. Lastly, the resulting RDD from the previous stage is joined with the scammers RDD to filter out the addresses that were not recognized as scammers to yield the addresses used by scammers to accumulate stolen Ether.

Results:

The results can be accessed via scammers.txt. A subset of results for refference: