If the data itself is random then it is an equally likely to the text space, so any transformation will lead safety. XOR or stream cipher maintains the relation that the blocks have in cipher-text as below..
from Crypto.Cipher import ARC4
key = '1234567812345678'
obj1 = ARC4.new(key)
obj2 = ARC4.new(key)
d1= obj1.encrypt('\x01\x82\x83\x04\x05\x06\x10\x81\x23\x32\x33\x34')
d2= obj2.encrypt('\x81\x02\x83\x84\x85\x86\x90\x01\xa3\xb2\x33\xb4')
print repr(d1)
print repr(d2)
p1='';p2=''
for i in d1:
if ord(i)>=128: p1+=chr(ord(i)-128)
else: p1+=chr(ord(i))
print; print
for i in d2:
if ord(i)>=128: p2+= chr(ord(i)-128)
else: p2+=chr(ord(i))
print p1==p2`
output:
'\xbaq\xba\xd0\x0c\xb7\xce&\xd3\x019\xfb'
':\xf1\xbaP\x8c7N\xa6S\x819{'
True