tlva.py

#!/usr/bin/env python3

#  leakfinder (tlva.py)
# -----------------
#  -f <traceset>
#  -d <distinguisher>
#     - even
#     - fixed
#     - random
#     - (an attack name), same as dpa
#  -b <bytenum:default 0>
#  -k <keyguess:default 0>>
#  --show-variance: show intra-group variance (glitchfinder)

import scipy
import scipy.stats
import getopt
import random
import sys
import numpy as np
import warnings
import matplotlib as mpl
import sparkgap.filemanager
import sparkgap.attack
import time

def getHammingWeight(x):
  return bin(x).count("1")

print("Initializing HW LUT")
HW_LUT = [getHammingWeight(x) for x in range(0,256)]

CONFIG_FILE = None

lastTime = 0.0
lastX = 0

def onclick(event):
  global lastTime, lastX
  t = time.time()
  if t - lastTime < 0.200:
    print("debounce - nope")
    return
  elif event.xdata is None:
    print("skip - event.xdata (click on graph) is none")
    return
  else:
    lastTime = t
    if lastX == 0:
      lastX = int(event.xdata)
      print("MARK: %d" % lastX)
    else:
      localX = int(event.xdata)
      fromX = min(lastX,localX)
      toX = max(lastX,localX)
      dist = toX - fromX
      print("FROM %d TO %d DIST %d" % (fromX,toX,dist))
      lastX = localX

def distinguisher_fixed(data,ct,round,keyguess):
  return np.array_equal(data,[0xaa] * 16)

def distinguisher_even(data,ct,round,keyguess):
  # print("%02x" % data[round])
  return data[round] % 2 == 0

def distinguisher_lsb2(data,ct,round,keyguess):
  # print("%02x" % data[round])
  return (data[round] >> 1) % 2 == 0

def distinguisher_hw(data,ct,round,keyguess):
  return HW_LUT[data[round]] >= 4

# def distinguisher_hw3232323232323232323232323232323232323232323232323232323232323232(data,ct,round,keyguess):
def distinguisher_hw32(data,ct,round,keyguess):
  out = 0 
  for i in range(0,4):
    out += HW_LUT[data[round * 4 + i]]
  return out >= 16

def distinguisher_hw64(data,ct,round,keyguess):
  out = 0 
  for i in range(0,8):
    out += HW_LUT[data[round * 4 + i]]
  return out >= 32

def distinguisher_cthw(data,ct,round,keyguess):
  return HW_LUT[ct[round]] >= 4

def distinguisher_hw_cryp(data,ct,round,keyguess):
  bc = 0
  for i in range(0,4):
    bc += HW_LUT[data[i]]
  return bc >= 16

# keeloq is 66 bits...
def unpackKeeloq(plaintext):
  out = ""
  for i in range(0,9):
    out = format(plaintext[i],"08b") + out
  return out[6:]

def distinguisher_keeloq(data,round):
  f = unpackKeeloq(data)
  # print(f)
  fc = f[0:32]
  return fc[round] == '1'

def distinguisher_random(data,ct,round,keyguess):
  return random.randint(0,10) % 2 == 0

def distinguisher_pt32(pt,ct,round,key):
  # return out
  out = 0
  for i in range(round * 4 + 0,round * 4 + 4):
    out += HW_LUT[pt[i]]
  return out >= 16

def distinguisher_ct32(pt,ct,round,key):
  # return out
  out = 0
  for i in range(round * 4 + 0,round * 4 + 4):
    out += HW_LUT[ct[i]]
  return out >= 16

CONFIG_DISTINGUISHER = distinguisher_fixed

# plaintext dpa "t-test"
def do_dpatest(fn,leakmodel,round,CONFIG_BYTE,CONFIG_KEYGUESS):
  tlva_grp1 = []
  tlva_grp2 = []
  for tracenum in range(0,fn.traceCount):
    if leakmodel.distinguisher(tracenum,CONFIG_BYTE,CONFIG_KEYGUESS):
      tlva_grp1.append(fn.getSingleTrace(tracenum))
    else:
      tlva_grp2.append(fn.getSingleTrace(tracenum))
  print("Group 1 count: %d" % len(tlva_grp2))
  print("Group 2 count: %d" % len(tlva_grp1))
  tlva_grp1_mean = np.mean(tlva_grp1,axis=0)
  tlva_grp2_mean = np.mean(tlva_grp2,axis=0)
  return np.abs(tlva_grp1_mean - tlva_grp2_mean)  

 
def do_tlva(fn,leakmodel,round,CONFIG_BYTE,CONFIG_KEYGUESS):
  cf = [0xAA] * 16
  tlva_fixed_traces = []
  tlva_random_traces = []
  for tracenum in range(0,fn.traceCount):
    if leakmodel.distinguisher(tracenum,CONFIG_BYTE,CONFIG_KEYGUESS):
      tlva_fixed_traces.append(fn.getSingleTrace(tracenum))
    else:
      tlva_random_traces.append(fn.getSingleTrace(tracenum))
  print("Fixed traces count: %d" % len(tlva_fixed_traces))
  print("Random traces count: %d" % len(tlva_random_traces))
  if len(tlva_fixed_traces) == 0:
    print("Padding fixed traces...")
    tlva_fixed_traces = np.array([[np.nan for _ in range(fn.numPoints)]])
  with warnings.catch_warnings():
    warnings.simplefilter("ignore")
    ttrace = scipy.stats.ttest_ind(tlva_random_traces,tlva_fixed_traces,axis=0,equal_var=False)
  variance_group1 = scipy.stats.variation(tlva_fixed_traces)
  variance_group2 = scipy.stats.variation(tlva_random_traces)
  return (np.nan_to_num(ttrace[0]),np.nan_to_num(ttrace[1]),fn.numPoints,variance_group1,variance_group2)

CONFIG_WRITEFILE = None

DO_TLVA  = 0
DO_DPA   = 1
CONFIG_STRATEGY = DO_TLVA
SHOW_VARIANCE = False

leakmodel = None

class SpecialLeakModel:
  def __init__(self):
    self.customDistinguisher = None
    pass

  def loadPlaintextArray(self,pts):
    self.plaintexts = pts

  def loadCiphertextArray(self,cts):
    self.ciphertexts = cts

  def distinguisher(self,xindex,byte,key):
    return self.customDistinguisher(self.plaintexts[xindex],self.ciphertexts[xindex],byte,key)

OptionManager = {}
SpecialDistinguisherList = {}
SpecialDistinguisherList["EVEN"] = (distinguisher_even,"Even and odd, first byte of plaintext")
SpecialDistinguisherList["LSB2"] = (distinguisher_lsb2,"LSB2 even or odd, first byte of plaintext")
SpecialDistinguisherList["HW"] = (distinguisher_hw,"HW >= 4, first byte of plaintext")
SpecialDistinguisherList["HW32"] = (distinguisher_hw32,"HW[0:4] >= 16, hamming weight 32-bit (first dword)")
SpecialDistinguisherList["CTHW"] = (distinguisher_cthw,"HW >= 4, first byte of ciphertext")
SpecialDistinguisherList["HW64"] = (distinguisher_hw64,"HW[0:8] >= 32, hamming weight 64-bit")
SpecialDistinguisherList["FIXED"] = (distinguisher_fixed,"Most common PT and all other PT's")
SpecialDistinguisherList["RANDOM"] = (distinguisher_random,"Randomly sorts two piles")
SpecialDistinguisherList["PT32"] = (distinguisher_pt32,"HW >= 16, first dword of plaintext")
SpecialDistinguisherList["CT32"] = (distinguisher_ct32,"HW >= 16, first dword of ciphertext")

CONFIG_SAMPLEOFFSET = None
CONFIG_SAMPLECOUNT = None

if __name__ == "__main__":
  optlist, args = getopt.getopt(sys.argv[1:],"f:d:w:r:b:k:a:o:n:h",["help","byte=","key=","distinguisher=","writefile=","round=","dpa","opt=","show-variance","attack="])
  CONFIG_ROUND = 0
  CONFIG_BYTE = 0
  CONFIG_KEY = 0
  for arg, value in optlist:
    if arg == "-f":
      CONFIG_FILE = value
    elif arg in ["--byte","-b"]:
      CONFIG_BYTE = int(value,0)
    elif arg in ["--key","-k"]:
      CONFIG_KEY = int(value,0)
    elif arg == "-o":
      CONFIG_SAMPLEOFFSET = int(value,0)
    elif arg == "-n":
      CONFIG_SAMPLECOUNT = int(value,0)
    elif arg == "--opt":
      try:
        (key,val) = value.split(":")
        OptionManager[key.strip()] = val.strip()
      except:
        print("Fatal: could not split '%s' on ':'" % arg)
        sys.exit(0)
    elif arg == "--show-variance":
      SHOW_VARIANCE = True
    elif arg == "--dpa":
      print("* Using raw plaintext DPA technique")
      CONFIG_STRATEGY = DO_DPA
    elif arg in ["-h","--help"]:
      print("usage: ./tlva.py -f [file] -d [distinguisher]")
      print("supported custom distinguishers:")
      for k in SpecialDistinguisherList.keys():
        print("  %s : %s" % (k,SpecialDistinguisherList[k][1]))
      sys.exit(0)
    elif arg in ["-r","--round"]:
      CONFIG_ROUND = int(value,0)
    elif arg in ["-d","--distinguisher","-a","--attack"]:
      # CONFIG_STRATEGY = DO_DPA
      if value.upper() in SpecialDistinguisherList.keys():
        print("Using special distinguisher '%s'" % value.upper())
        leakmodel = SpecialLeakModel()
        (d,t) = SpecialDistinguisherList[value.upper()]
        leakmodel.customDistinguisher = d
      else:
        print("Attempting to use distinguisher '%s'" % value)
        leakmodel = sparkgap.attack.fetchModel(value)
    elif arg in ["-w","--writefile"]:
      CONFIG_WRITEFILE = value
  if CONFIG_WRITEFILE is not None:
    mpl.use("Agg")  
  import matplotlib.pyplot as plt
  if CONFIG_FILE is None:
    print("You must specify a file")
    sys.exit(0)
  fn = sparkgap.filemanager.TraceManager(CONFIG_FILE)
  if hasattr(leakmodel,"loadOptions"):
    print("Loading options for leakmodel")
    leakmodel.loadOptions(OptionManager)
  if leakmodel is None:
    print("You must specify a leak model")
    sys.exit(0)
  leakmodel.loadPlaintextArray(fn.loadPlaintexts())
  leakmodel.loadCiphertextArray(fn.loadCiphertexts())
  if CONFIG_STRATEGY == DO_TLVA:
    (tt,tp,numSamples,var_group1,var_group2) = do_tlva(fn,leakmodel,CONFIG_ROUND,CONFIG_BYTE,CONFIG_KEY)
  elif CONFIG_STRATEGY == DO_DPA:
    dpa_diff = do_dpatest(fn,leakmodel,CONFIG_ROUND,CONFIG_BYTE,CONFIG_KEY)
    print(dpa_diff)
    # needs different graph mechanism.
    fig,(ax1,ax2) = plt.subplots(2,1)
    if CONFIG_SAMPLECOUNT is not None and CONFIG_SAMPLEOFFSET is not None:
      ax1.plot(fn.getSingleTrace(0)[CONFIG_SAMPLEOFFSET:CONFIG_SAMPLECOUNT+CONFIG_SAMPLEOFFSET],color="grey")
      ax2.plot(dpa_diff[CONFIG_SAMPLEOFFSET:CONFIG_SAMPLECOUNT + CONFIG_SAMPLEOFFSET],color="red")
    else:
      ax1.plot(fn.getSingleTrace(0),color="grey")
      ax2.plot(dpa_diff,color="red")
    plt.show()
    sys.exit(0)
  else:
    print("Unknown value of CONFIG_STRATEGY, %d" % CONFIG_STRATEGY)
    sys.exit(0)
  if SHOW_VARIANCE:
    fig,(ax1,ax2,ax3) = plt.subplots(3,1)
    ax2.set_title("Variance, Group 1")
    ax2.plot(var_group1)
    ax3.set_title("Variance, Group 2")
    ax3.plot(var_group2)
  else:
    fig,ax1 = plt.subplots()
  fig.canvas.mpl_connect("button_press_event",onclick)
  fig.canvas.manager.set_window_title("Test Vector Leakage Assessment")
  ax1.set_title("T-Value")
  ax1.set_xlabel("Sample")
  ax1.set_ylabel("T-Test Value")
  t = np.argmax(abs(tt))
  print("Argmax is %d, tvalue is %f" % (t,tt[t]))
  ax1.plot(tt)
  ax1.plot(t,tt[t],marker='x')
  tlen = numSamples
  if CONFIG_STRATEGY == DO_TLVA:
    print("Drawing 4.5 markers")
    ax1.hlines(y=4.5,xmin=0,xmax=tlen,color='r')
    ax1.hlines(y=-4.5,xmin=0,xmax=tlen,color='r')
  if CONFIG_WRITEFILE is not None:
    plt.savefig(CONFIG_WRITEFILE)
  else:
    plt.show()