Scripts in the bitcoin protocol

There is a point that we have touched upon several times but not yet properly explained – the role of scripts in the bitcoin protocol. We have seen that the public key and the signature are stored inside a bitcoin transaction in container data structures that were called scriptPubKey and scriptSig in the source code. But scripts are more than just container for data, they are in fact executable and do not only store keys and signatures but also instruct the bitcoin server what to do with this data in order to validate a signature.

So what is a script?

In fact, a script is a combination of instructions, called opcodes, and data. During execution, the data is placed on a stack. Instructions can remove (“pop”) data from the top of the stack, operate on them and then place a result on the stack (“push”). If you have ever seen Forth, that might sound familiar.

Let me guide you through the relevant source code in the reference implementation to give you an idea what role scripts play in the bitcoin protocol. The key location in the code is the function VerifyScript in script/interpreter.cpp. This function is called when the bitcoin server needs to validate a transaction input, for instance before adding it to a block so that it becomes part of the blockchain or upon receiving it from a peer in the network. In this case, this function will be called, passing it – among other parameters – the signature script that is part of the transaction input, the public key script that is part of the transaction output referenced by that input and – via an instance of the class BaseSignatureChecker – a reference to the transaction of which the transaction input is a part.

The key sequence in this function is the following code snippet.

std::vector stack, stackCopy;
if (!EvalScript(stack, scriptSig, flags, checker, SIGVERSION_BASE, serror))
    // serror is set
    return false;
if (flags & SCRIPT_VERIFY_P2SH)
    stackCopy = stack;
if (!EvalScript(stack, scriptPubKey, flags, checker, SIGVERSION_BASE, serror))
    // serror is set
    return false;
if (stack.empty())
    return set_error(serror, SCRIPT_ERR_EVAL_FALSE);
if (CastToBool(stack.back()) == false)
    return set_error(serror, SCRIPT_ERR_EVAL_FALSE);

So basically the following things happen. First, a clean stack – implemented as a vector – is created. Then, the signature script is evaluated and operates on this stack. Next, the public key script of the spent transaction output is executed, using the same stack. If, after executing both scripts, the stack is either empty or the element at the top is not the boolean value True, the verification has failed.

Now, in most cases, both the signature script and the public key script are not arbitrary, but follow a small number of defined patterns that, when executed, are effectively equivalent to validating the signature of the transaction (these patterns are defined in the function Solver in script/standard.cpp. The most commonly used of these patterns is called Pay to public key hash (P2PKH). In this case, the signature script does in fact not contain any real opcodes, but consists of two pieces of data that are pushed on the stack. Here and in the sequel, we will use capital letters to denote a script instruction and data in brackets to denote data. The first piece of data that is pushed on the stack is the signature of the transaction. The second piece is a hash of the public key that belongs to the private key which was used to make the signature. If, for instance, Alice is building a transaction to transfer bitcoin to Bob, this will be the public key of Alice. Thus a P2PKH signature script looks as follows.

[signature] [public key]

After that script has executed, we will therefore find two items on the stack. At the top of the stack, there will be the public key, and below, at the bottom of the stack, there will be the signature.

[public key]
[signature]

The next script that the engine will evaluate is the public key script. In the case of a standard P2PKH script, this is a bit more complicated and looks as follows.

OP_DUP OP_HASH160 [public key hash] OP_EQUALVERIFY OP_CHECKSIG

Let us go through the instructions one by one. To understand how each of the opcodes is processed, you will have to look at the code in EvalScript in script/interpreter.cpp. The first opcode that is executed is OP_DUP. This command will simply duplicate the item at the top of the stack. As we are still working with the stack left over after executing the signature script, our stack will look as follows after the OP_DUP has been executed.

[public key]
[public key]
[signature]

The next command is OP_HASH160. This instruction will pop one item off the stack, will calculate its 160 bit hash value (SHA2560 followed by RIPEMD160) and push the result back onto the stack. Thus our stack now is

[public key hash]
[public key]
[signature]

The next command is again not really a command, but just pushes the value on the stack. Thus after executing this part of the script, the stack now contains four items.

[public key hash]
[public key hash]
[public key]
[signature]

The next instruction is the instruction OP_EQUALVERIFY. This operation removes the two top items from the stack, compares them and raises an error if the two values are not equal. Now remember where these two items came from. One of them originates from the transaction input – it was the result of taking the hash value of the public key that Alice added to the transaction input along with the signature, i.e. the public key matching her private key. The other one was part of the spent transaction output. Thus this check fails unless – as it should be – the transaction output has been paid to Alice’s public key hash, so that these two items are equal.

If that is the case and we survive this verification step, two of the four stack items will have been gone and our stack is

[public key]
[signature]

Now the last instruction is executed. This instruction – OP_CHECKSIG – removes the two items left (the public key and the signature) from the stack and actually validates the signature, similar to what we have done in my post on elliptic curve cryptography. If the signature could be successfully validated, a boolean “True” is pushed on the stack, and the verification of the script completes successfully, otherwise “False” is pushed and the script verification returns an error.

Thus, effectively, executing this script amounts to comparing the public key in the transaction input with the public key hash in the spent transaction output and verifying the signature contained in the transaction input.

An example

Of course, in a bitcoin transaction, the scripts are not included in a nice, readable way as we have presented them here. Instead, a script is simply a sequence of bytes. Time to take a look at an example in depth. In one of the last posts, we looked at the transaction ed70b8c66a4b064cfe992a097b3406fa81ff09641fe55a709e4266167ef47891. Let us now analyse this transaction in detail. The following ipython session assumes that you have downloaded my btc library.

In [1]: import btc.utils
In [2]: raw = btc.utils.getRawTransaction(txid="ed70b8c66a4b064cfe992a097b3406fa81ff09641fe55a709e4266167ef47891")
In [3]: import btc.txn
In [4]: txn = btc.txn.txn()
In [5]: txn.deserialize(raw)
In [6]: txin = txn.getInputs()[0]
In [7]: script = txin.getScriptSigHex()
In [8]: script
Out[8]: '47304402207f5dfc2f7f7329b7cc731df605c83aa6f48ec2218495324bb4ab43376f313b840220020c769655e4bfcc54e55104f6adc723867d9d819266d27e755e098f646f689d0121038c2d1cbe4d731c69e67d16c52682e01cb70b046ead63e90bf793f52f541dafbd'

To understand how this script is translated into opcodes, we need to look at CScript::GetOpcode() in script/script.h. Here we find that any number x smaller than OP_PUSHDATA1, i.e. 0x4c, is interpreted as an instruction to push the following x bytes onto the stack – this is called an immediate push. If we match this with the general description of a signature script in the P2PKH standard above, we find that the next 0x47 = 71 bytes are the private key.

In [9]: sig = txin.getScriptSigHex()[2:2*71+2]
In [10]: sig
Out[10]: '304402207f5dfc2f7f7329b7cc731df605c83aa6f48ec2218495324bb4ab43376f313b840220020c769655e4bfcc54e55104f6adc723867d9d819266d27e755e098f646f689d01'
In [11]: script = txin.getScriptSigHex()[2*71+2:]
In [12]: script
Out[12]: '21038c2d1cbe4d731c69e67d16c52682e01cb70b046ead63e90bf793f52f541dafbd'

Looking at the remaining part of the script, we see that the next opcode is again an immediate push, pushing the remaining part of the script onto the stack. So we can conclude that this remaining part must be the public key.

In [13]: pub = script[2:]
In [14]: pub
Out[14]: '038c2d1cbe4d731c69e67d16c52682e01cb70b046ead63e90bf793f52f541dafbd'

We see that the public key starts with 0x03, so it is a compressed key, and we could encode it using our findings from my previous post on this.

The signature, however, looks a bit more mysterious, so let us try to understand the content of the variable sig. This is a so called DER encoded signature, with an additional byte appended at the end.

The DER standard is an ASN.1 encoding standard which is part of the X.690 specification. In the source code, the format is described in IsValidSignatureEncoding in script/interpreter.cpp. Essentially a DER encoded signature is mixture of structural information like data types and lengths, and the actual data, i.e. the integers r and s that make up an ECDSA signature. If you need the details, you will find a Python implementation of a decoding routine here.

Let us now take a look at the spent transaction where we find the second part of the script.

In [15]: prev_raw = btc.utils.getRawTransaction(txid=txin.getPrevTxId())
In [16]: prev = btc.txn.txn()
In [17]: prev.deserialize(prev_raw)
In [18]: index = txin.getVout()
In [19]: spentTxo = prev.getOutputs()[index]
In [20]: script = spentTxo.getScriptPubKeyHex()
In [21]: script
Out[21]: '76a914140268d5d1c4e1792db22e4776edf3c168fd59f588ac'

Let us again try to translate this script into transactions and data. In script.h, we find that the first byte is the opcode OP_DUP = 0x76. The second byte is the opcode OP_HASH160 = 0xa9. The third byte is again an immediate push and will push 0x14 = 20 bytes onto the stack, which is the public key hash.

In [22]: script = script[6:]
In [23]: pubKeyHash = script[:40]
In [24]: script = script[40:]
In [25]: script
Out[25]: '88ac'

Thus we are left with the two bytes 0x88 and 0xac. These bytes are again opcodes, namely OP_EQUALVERIFY and OP_CHECKSIG. So we find that our public key script has exactly the form that we expect.

We have now dissected a serialized transaction down to the last byte and have looked at how the process of verifying the signature is decoded as a combination of a signature script and a public key script (which are sometimes called the unlocking and locking script). In one of the next posts, we will look at the process of creating a signature in more detail and we will assemble a full bitcoin transaction which we can push into our local test network.

One thought on “Scripts in the bitcoin protocol

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s