acsm-calibre-plugin/calibre-plugin/libadobe.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

'''
Helper library with code needed for Adobe stuff.
'''

from uuid import getnode
import os, hashlib, base64
import urllib.request, ssl
from datetime import datetime, timedelta

from lxml import etree
import rsa

try:
    from Crypto import Random
    from Crypto.Cipher import AES
    from Crypto.PublicKey import RSA
    from Crypto.Hash import SHA
except ImportError:
    # Debian (and Ubuntu) ship pycryptodome, but not in its compatible mode with pycrypto
    # If `Crypto` can't be found, try under pycryptodome's own namespace
    from Cryptodome import Random
    from Cryptodome.Cipher import AES
    from Cryptodome.PublicKey import RSA
    from Cryptodome.Hash import SHA

from oscrypto import keys
from oscrypto.asymmetric import dump_certificate, dump_private_key, dump_public_key



VAR_ACS_SERVER_HTTP = "http://adeactivate.adobe.com/adept"
VAR_ACS_SERVER_HTTPS = "https://adeactivate.adobe.com/adept"

FILE_DEVICEKEY = "devicesalt"
FILE_DEVICEXML = "device.xml"
FILE_ACTIVATIONXML = "activation.xml"


# Lists of different ADE "versions" we know about
VAR_VER_SUPP_CONFIG_NAMES = [ "ADE 1.7.2", "ADE 2.0.1", "ADE 3.0.1", "ADE 4.0.3", "ADE 4.5.10", "ADE 4.5.11" ]
VAR_VER_SUPP_VERSIONS = [ "ADE WIN 9,0,1131,27", "2.0.1.78765", "3.0.1.91394", "4.0.3.123281", 
                            "com.adobe.adobedigitaleditions.exe v4.5.10.186048", 
                            "com.adobe.adobedigitaleditions.exe v4.5.11.187303" ]
VAR_VER_HOBBES_VERSIONS = [ "9.0.1131.27", "9.3.58046", "10.0.85385", "12.0.123217", "12.5.4.186049", "12.5.4.187298" ]
VAR_VER_OS_IDENTIFIERS = [ "Windows Vista", "Windows Vista", "Windows 8", "Windows 8", "Windows 8", "Windows 8" ]


# "Missing" versions:
# 1.7.1, 2.0, 3.0, 4.0, 4.0.1, 4.0.2, 4.5 to 4.5.9

# This is a list of ALL versions we know (and can potentially use if present in a config file).
# Must have the same length / size as the four lists above.
VAR_VER_BUILD_IDS = [ 1131, 78765, 91394, 123281, 186048, 187303 ]
# Build ID 185749 also exists, that's a different (older) variant of 4.5.10. 

# This is a list of versions that can be used for new authorizations:
VAR_VER_ALLOWED_BUILD_IDS_AUTHORIZE = [ 78765, 91394, 123281, 187303 ]

# This is a list of versions to be displayed in the version changer.
VAR_VER_ALLOWED_BUILD_IDS_SWITCH_TO = [ 1131, 78765, 91394, 123281, 187303 ]

# Versions >= this one are using HTTPS
# According to changelogs, this is implemented as of ADE 4.0.1 - no idea what build ID that is.
VAR_VER_NEED_HTTPS_BUILD_ID_LIMIT = 123281

# Default build ID to use - ADE 2.0.1
VAR_VER_DEFAULT_BUILD_ID = 78765



def are_ade_version_lists_valid():
    # These five lists MUST all have the same amount of elements. 
    # Otherwise that will cause all kinds of issues. 

    fail = False
    if len(VAR_VER_SUPP_CONFIG_NAMES) != len(VAR_VER_SUPP_VERSIONS):
        fail = True
    if len(VAR_VER_SUPP_CONFIG_NAMES) != len(VAR_VER_HOBBES_VERSIONS):
        fail = True
    if len(VAR_VER_SUPP_CONFIG_NAMES) != len(VAR_VER_OS_IDENTIFIERS):
        fail = True
    if len(VAR_VER_SUPP_CONFIG_NAMES) != len(VAR_VER_BUILD_IDS):
        fail = True
    
    if fail:
        print("Internal error in DeACSM: Mismatched version list lenghts.")
        print("This should never happen, please open a bug report.")
        return False
    
    return True


devkey_bytes = None



def get_devkey_path():
    global FILE_DEVICEKEY
    return FILE_DEVICEKEY
def get_device_path():
    global FILE_DEVICEXML
    return FILE_DEVICEXML
def get_activation_xml_path():
    global FILE_ACTIVATIONXML
    return FILE_ACTIVATIONXML


def update_account_path(folder_path: str):
    global FILE_DEVICEKEY, FILE_DEVICEXML, FILE_ACTIVATIONXML

    FILE_DEVICEKEY = os.path.join(folder_path, "devicesalt")
    FILE_DEVICEXML = os.path.join(folder_path, "device.xml")
    FILE_ACTIVATIONXML = os.path.join(folder_path, "activation.xml")


def createDeviceKeyFile():
    # Original implementation: Device::createDeviceKeyFile()

    DEVICE_KEY_SIZE = 16
    global devkey_bytes
    devkey_bytes = Random.get_random_bytes(DEVICE_KEY_SIZE)

    f = open(FILE_DEVICEKEY, "wb")
    f.write(devkey_bytes)
    f.close()

def get_mac_address(): 
    mac1 = getnode()
    mac2 = getnode()
    if (mac1 != mac2) or ((mac1 >> 40) % 2):
        return bytes([1, 2, 3, 4, 5, 0])
    
    return mac1.to_bytes(6, byteorder='big')



def makeSerial(random: bool):
    # Original implementation: std::string Device::makeSerial(bool random)

    # It doesn't look like this implementation results in the same fingerprint Adobe is using in ADE.
    # Given that Adobe only ever sees the SHA1 hash of this value, that probably doesn't matter.

    sha_out = None

    if not random: 
        try:
            # Linux
            uid = os.getuid()
            import pwd
            username = pwd.getpwuid(uid).pw_name.encode("utf-8").decode("latin-1")
        except: 
            # Windows
            uid = 1000
            username = os.getlogin().encode("utf-8").decode("latin-1")

        mac_address = get_mac_address()

        dataToHash = "%d:%s:%02x:%02x:%02x:%02x:%02x:%02x\x00" % (uid, username, 
            mac_address[0], mac_address[1], mac_address[2], 
            mac_address[3], mac_address[4], mac_address[5])

        sha_out = hashlib.sha1(dataToHash.encode('latin-1')).hexdigest().lower()
    else: 
        sha_out = Random.get_random_bytes(20).hex().lower()

    return sha_out

def makeFingerprint(serial: str):
    # Original implementation: std::string Device::makeFingerprint(const std::string& serial)
    # base64(sha1(serial + privateKey))
    # Fingerprint must be 20 bytes or less.

    f = open(FILE_DEVICEKEY, "rb")
    devkey_bytes = f.read()
    f.close()

    str_to_hash = serial + devkey_bytes.decode('latin-1')
    hashed_str = hashlib.sha1(str_to_hash.encode('latin-1')).digest()
    b64str = base64.b64encode(hashed_str)

    return b64str


############################################## HTTP stuff:

def sendHTTPRequest_DL2FILE(URL: str, outputfile: str):
    headers = {
        "Accept": "*/*",
        "User-Agent": "book2png",
        # MacOS uses different User-Agent. Good thing we're emulating a Windows client.
    }
    req = urllib.request.Request(url=URL, headers=headers)
    handler = urllib.request.urlopen(req)

    chunksize = 16 * 1024

    ret_code = handler.getcode()


    loc = None
    try: 
        loc = req.headers.get("Location")
    except:
        pass

    if loc is not None: 
        return sendHTTPRequest_DL2FILE(loc)

    if ret_code != 200:
        return ret_code

    with open(outputfile, "wb") as f:
        while True: 
            chunk = handler.read(chunksize)
            if not chunk: 
                break
            f.write(chunk)

    return 200

def sendHTTPRequest_getSimple(URL: str):

    headers = {
        "Accept": "*/*",
        "User-Agent": "book2png",
        # MacOS uses different User-Agent. Good thing we're emulating a Windows client.
    }

    # Ignore SSL:
    # It appears as if lots of book distributors have either invalid or expired certs ...
    # No idea how Adobe handles that (pinning?), but we can just ignore SSL errors and continue anyways.
    # Not the best solution, but it works.
    ctx = ssl.create_default_context()
    ctx.check_hostname = False
    ctx.verify_mode = ssl.CERT_NONE

    req = urllib.request.Request(url=URL, headers=headers)
    handler = urllib.request.urlopen(req, context=ctx)

    content = handler.read()

    loc = None
    try: 
        loc = req.headers.get("Location")
    except:
        pass

    if loc is not None: 
        return sendHTTPRequest_getSimple(loc)

    return content

def sendPOSTHTTPRequest(URL: str, document: bytes, type: str, returnRC = False):

    headers = {
        "Accept": "*/*",
        "User-Agent": "book2png",
        # MacOS uses different User-Agent. Good thing we're emulating a Windows client.
        "Content-Type": type
    }

    # Ignore SSL:
    # It appears as if lots of book distributors have either invalid or expired certs ...
    # No idea how Adobe handles that (pinning?), but we can just ignore SSL errors and continue anyways.
    # Not the best solution, but it works.
    ctx = ssl.create_default_context()
    ctx.check_hostname = False
    ctx.verify_mode = ssl.CERT_NONE

    req = urllib.request.Request(url=URL, headers=headers, data=document)
    handler = urllib.request.urlopen(req, context=ctx)

    ret_code = handler.getcode()
    if (ret_code == 204 and returnRC):
        return 204, ""
    if (ret_code != 200):
        print("Post request returned something other than 200 - returned %d" % (ret_code))

    content = handler.read()

    loc = None
    try: 
        loc = req.headers.get("Location")
    except:
        pass

    if loc is not None: 
        return sendPOSTHTTPRequest(loc, document, type, returnRC)

    if returnRC:
        return ret_code, content

    return content


def sendHTTPRequest(URL: str):
    return sendHTTPRequest_getSimple(URL)


def sendRequestDocu(document: str, URL: str):
    return sendPOSTHTTPRequest(URL, document.encode("utf-8"), "application/vnd.adobe.adept+xml", False)

def sendRequestDocuRC(document: str, URL: str):
    return sendPOSTHTTPRequest(URL, document.encode("utf-8"), "application/vnd.adobe.adept+xml", True)



######### Encryption and signing ###################


def encrypt_with_device_key(data):

    global devkey_bytes
    if devkey_bytes is None: 
        f = open(FILE_DEVICEKEY, "rb")
        devkey_bytes = f.read()
        f.close()

    remain = 16
    if (len(data) % 16):
        remain = 16 - (len(data) % 16)

    data += bytes([remain])*remain

    iv = Random.get_random_bytes(16)
    cip = AES.new(devkey_bytes, AES.MODE_CBC, iv)
    encrypted = cip.encrypt(data)

    res = iv + encrypted
    return res

def decrypt_with_device_key(data): 
    global devkey_bytes
    if devkey_bytes is None: 
        f = open(FILE_DEVICEKEY, "rb")
        devkey_bytes = f.read()
        f.close()

    cip = AES.new(devkey_bytes, AES.MODE_CBC, data[:16])
    decrypted = cip.decrypt(data[16:])

    # Remove padding
    decrypted = decrypted[:-decrypted[-1]]

    return decrypted


def addNonce(): 

    dt = datetime.utcnow()
    usec = dt.microsecond
    sec = (dt - datetime(1970,1,1)).total_seconds()


    Ntime = int(int(sec * 1000) + usec/1000)

    # Unixtime to gregorian timestamp
    Ntime += 62167219200000

    final = bytearray(Ntime.to_bytes(8, 'little'))

    # Something is fishy with this tmp value. It usually is 0 in ADE, but not always. 
    # I haven't yet figured out what it means ...
    tmp = 0
    final.extend(tmp.to_bytes(4, 'little'))

    ret = ""

    ret += "<adept:nonce>%s</adept:nonce>" % (base64.b64encode(final).decode("utf-8"))

    m10m = dt + timedelta(minutes=10)
    m10m_str = m10m.strftime("%Y-%m-%dT%H:%M:%SZ")

    ret += "<adept:expiration>%s</adept:expiration>" % (m10m_str)

    return ret


def get_cert_from_pkcs12(_pkcs12, _key):

    _, cert, _ = keys.parse_pkcs12(_pkcs12, _key)
    cert = dump_certificate(cert, encoding="der")

    return cert


def sign_node(node):

    sha_hash = hash_node(node)
    sha_hash = sha_hash.digest()

    # print("Hash is " + sha_hash.hex())

    global devkey_bytes
    global pkcs12

    if devkey_bytes is None: 
        f = open(FILE_DEVICEKEY, "rb")
        devkey_bytes = f.read()
        f.close()

    try: 
        activationxml = etree.parse(FILE_ACTIVATIONXML)
        adNS = lambda tag: '{%s}%s' % ('http://ns.adobe.com/adept', tag)
        pkcs12 = activationxml.find("./%s/%s" % (adNS("credentials"), adNS("pkcs12"))).text
    except: 
        return None

    my_pkcs12 = base64.b64decode(pkcs12)
    my_priv_key, _, _ = keys.parse_pkcs12(my_pkcs12, base64.b64encode(devkey_bytes))
    my_priv_key = dump_private_key(my_priv_key, None, "der")

    key = rsa.PrivateKey.load_pkcs1(RSA.importKey(my_priv_key).exportKey())
    keylen = rsa.pkcs1.common.byte_size(key.n)
    padded = rsa.pkcs1._pad_for_signing(sha_hash, keylen)
    payload = rsa.pkcs1.transform.bytes2int(padded)
    encrypted = key.blinded_encrypt(payload)
    block = rsa.pkcs1.transform.int2bytes(encrypted, keylen)
    signature = base64.b64encode(block).decode()

    # print("sig is %s\n" % block.hex())

    return signature


 

def hash_node(node):

    hash_ctx = SHA.new()
    hash_node_ctx(node, hash_ctx)
    return hash_ctx



ASN_NONE = 0
ASN_NS_TAG = 1      # aka "BEGIN_ELEMENT"
ASN_CHILD = 2       # aka "END_ATTRIBUTES"
ASN_END_TAG = 3     # aka "END_ELEMENT"
ASN_TEXT = 4        # aka "TEXT_NODE"
ASN_ATTRIBUTE = 5   # aka "ATTRIBUTE"

debug = False

def hash_node_ctx(node, hash_ctx):

    qtag = etree.QName(node.tag)

    if (qtag.localname == "hmac" or qtag.localname == "signature"):
        if (qtag.namespace == "http://ns.adobe.com/adept"):
            # Adobe HMAC and signature are not hashed
            return
        else: 
            print("Warning: Found hmac or signature node in unexpected namespace " + qtag.namespace)

    hash_do_append_tag(hash_ctx, ASN_NS_TAG)

    if qtag.namespace is None: 
        hash_do_append_string(hash_ctx, "")
    else:
        hash_do_append_string(hash_ctx, qtag.namespace)
    hash_do_append_string(hash_ctx, qtag.localname)


    attrKeys = node.keys()

    # Attributes need to be sorted
    attrKeys.sort()
    # TODO Implement UTF-8 bytewise sorting:
    # "Attributes are sorted first by their namespaces and
    # then by their names; sorting is done bytewise on UTF-8
    # representations."

    for attribute in attrKeys: 
        # Hash all the attributes
        hash_do_append_tag(hash_ctx, ASN_ATTRIBUTE)

        # Check for element namespace and hash that, if present:
        q_attribute = etree.QName(attribute)

        # Hash element namespace (usually "")
        # If namespace is none, use "". Else, use namespace.
        hash_do_append_string(hash_ctx, "" if q_attribute.namespace is None else q_attribute.namespace)

        # Hash (local) name and value
        hash_do_append_string(hash_ctx, q_attribute.localname)
        hash_do_append_string(hash_ctx, node.get(attribute))

    hash_do_append_tag(hash_ctx, ASN_CHILD)

    if (node.text is not None):
        # If there's raw text, hash that.

        # This code block used to just be the following:
        #   hash_do_append_tag(hash_ctx, ASN_TEXT)
        #   hash_do_append_string(hash_ctx, node.text.strip())
        # though that only works with text nodes < 0x7fff.
        # While I doubt we'll ever encounter text nodes larger than 32k in
        # this application, I want to implement the spec correctly.
        # So there's a loop going over the text, hashing 32k chunks.

        text = node.text.strip()
        textlen = len(text)
        if textlen > 0:
            done = 0
            remaining = 0
            while True: 
                remaining = textlen - done
                if remaining > 0x7fff:
                    print("Warning: Hashing text node larger than 32k.")
                    print("This usually doesn't happen, and I'm not sure if this is implemented correctly.")
                    print("If you run into issues, please open a bug report.")
                    remaining = 0x7fff

                hash_do_append_tag(hash_ctx, ASN_TEXT)
                hash_do_append_string(hash_ctx, text[done:done+remaining]) 

                done += remaining
                if done >= textlen:
                    break

    for child in node: 
        # If there's child nodes, hash these as well.
        hash_node_ctx(child, hash_ctx)



    hash_do_append_tag(hash_ctx, ASN_END_TAG)



def hash_do_append_string(hash_ctx, string: str):

    str_bytes = bytes(string, encoding="utf-8")

    length = len(str_bytes)
    len_upper = int(length / 256)
    len_lower = int(length & 0xFF)

    hash_do_append_raw_bytes(hash_ctx, [len_upper, len_lower])
    hash_do_append_raw_bytes(hash_ctx, str_bytes)

def hash_do_append_tag(hash_ctx, tag: int):

    if (tag > 5):
        return
    
    hash_do_append_raw_bytes(hash_ctx, [tag])

def hash_do_append_raw_bytes(hash_ctx, data: bytes):
    hash_ctx.update(bytearray(data))