def lastor_redistr(client_coord, fp_to_bw, fp_to_coord, threshold):
"""
Redistributes LASTor guard probabilities.
threshold = selection probability to fraction of the cost of the network
contributed
"""
if threshold < 1:
print("Error: threshold must be at least 1.")
return None
total_cost = 0
for fp, bw in fp_to_bw.items():
total_cost += relays.get_cost(bw)
thresh = threshold / total_cost
ratios = dict.fromkeys(fp_to_bw, 0)
excess = 1
curr_fps = list(fp_to_coord.keys())
cnt = 1
while excess > 0:
print(f'----- Round {cnt} -----')
cnt += 1
if len(curr_fps) == 0:
break
curr_fp_to_coord = {fp: fp_to_coord[fp] for fp in curr_fps}
curr_distr = lastor.compute_lt_selection_probs(client_coord,
curr_fp_to_coord)
print(f'sum distr (~1): {sum(curr_distr.values())}')
norm = excess
print(f'norm: {norm}')
num_over = 0
excess = 0
for fp in list(curr_fps):
cost = relays.get_cost(fp_to_bw[fp])
ratio = curr_distr[fp] / cost * norm
ratios[fp] += ratio
if ratios[fp] > thresh:
excess += ((ratios[fp] - thresh) * cost)
ratios[fp] = thresh
curr_fps.remove(fp)
num_over += 1
print(f'num over thresh: {num_over}')
redistr = {}
for fp, val in ratios.items():
prob = val * relays.get_cost(fp_to_bw[fp])
redistr[fp] = prob
return redistr
def disp_targeted_split_table(num_relays, bw_resource):
"""
Prints table showing IP splitting effects on targeted attack
against client in NYC (40.6943, -73.9249).
Format:
Relay # |relCost | vanilla prob | cumulative targ prob | coordinate
"""
print("Targeted IP splitting")
client_coord = (40.6943, -73.9249)
# emp_prob = 0
mal_coords_lst = []
fname = "attack_split_targ/malguardprobs.json"
res = json.load(open(fname, 'r'))
for coord, prob in res.items():
lat = float(coord[1:-1].split(',')[0])
lon = float(coord[1:-1].split(',')[1])
mal_coords_lst.append((lat, lon))
v_distr, v_prob = vanilla.compute_vanilla_guard_distr(guard_to_bw, bw_resource)
for i in range(0, num_relays):
cost = num_relays * relays.get_cost(bw_resource / num_relays)
tot_cost = sum(guard_to_cost.values()) + cost
rel_cost = cost / tot_cost
# untargeted with 1 client in client_lst
res = untargeted_prob([client_coord], mal_coords_lst[:i+1], bw_resource)
print(f"{i+1} | relCost: {rel_cost} | VT prob: {v_prob} | targ LT prob: {res[0]} | {mal_coords_lst[i]}")
return mal_coords_lst
def disp_untargeted_split_table(num_relays, bw_resource):
"""
Prints table showing IP splitting effects on untargeted attack.
Format:
Relay # |relCost | vanilla prob | avg untarg prob | max untarg prob with client coord | best mal coordinate
"""
print("Untargeted IP splitting")
client_lst = json.load(open("../data/geoclients200.json"))
mal_coords_lst = []
fname = "attack_split_untarg/malguardprobs.json"
res = json.load(open(fname, 'r'))
for coord, prob in res.items():
lat = float(coord[1:-1].split(',')[0])
lon = float(coord[1:-1].split(',')[1])
mal_coords_lst.append((lat, lon))
v_distr, v_prob = vanilla.compute_vanilla_guard_distr(guard_to_bw, bw_resource)
for i in range(0, num_relays):
cost = num_relays * relays.get_cost(bw_resource / num_relays)
tot_cost = sum(guard_to_cost.values()) + cost
rel_cost = cost / tot_cost
res = untargeted_prob(client_lst, mal_coords_lst[:i+1], bw_resource)
print(f"{i+1} | relCost: {rel_cost} | VT prob: {v_prob} | avg prob: {res[0]} | max prob: {res[2]} |{mal_coords_lst[i]}")
return mal_coords_lst
def counterraptor_init_ratios(client_as, fp_to_bw, fp_to_as, fp_to_cost):
"""
Returns selection probability to cost ratio for client_as
before applying redistribution algorithm to Counter-RAPTOR.
"""
client_to_all_res = json.load(
open("../counterraptor/all_reachable_resilience.json"))
client_to_guard_res = json.load(
open("../counterraptor/cg_resilience.json"))
alpha = 0.5
sample_size = 0.1 * len(fp_to_bw)
total_cost = sum(fp_to_cost.values())
ratios = dict.fromkeys(fp_to_bw, 0)
distr = counterraptor.compute_cr_selection_probs(client_as, fp_to_bw,
fp_to_as,
client_to_all_res,
client_to_guard_res,
alpha, sample_size)
for fp, prob in distr.items():
cost = relays.get_cost(fp_to_bw[fp])
ratio = prob * total_cost / cost
ratios[fp] = ratio
return ratios
def disp_untargeted_split_table_avg(client_as_lst, best_as_untargeted,
bw_resource, num_relays_lst):
"""
Prints table showing varying relays effect on untargeted attack
(average success).
"""
v_guard_probs, v_prob = vanilla.compute_vanilla_guard_distr(
guard_to_bw, bw_resource)
for num_relays in num_relays_lst:
sum_probs = 0
for client_as in client_as_lst:
all_guard_probs, mal_guard_prob = cr.compute_cr_guard_distr(
client_as, best_as_untargeted, client_to_all_res,
client_to_guard_res, guard_to_bw, ip_to_as, num_relays,
bw_resource, alpha, sample_size)
sum_probs += mal_guard_prob
mal_prob = sum_probs / len(client_as_lst)
cost = num_relays * relays.get_cost(bw_resource / num_relays)
tot_cost = sum(guard_to_cost.values()) + cost
rel_cost = cost / tot_cost
print(
f"{num_relays} | relCost: {rel_cost} | VT prob: {v_prob} | utarg CR prob: {mal_prob} | AS{best_as_untargeted}"
)
def disp_max_advantages(client_as_lst, bw_resources_lst, best_as_dict):
"""
Display maximum attacker advantage in untargeted attack compared to
Vanilla Tor and relCost.
"""
for i in range(0, len(bw_resources_lst)):
bw_resource = bw_resources_lst[i]
v_guard_probs, v_prob = vanilla.compute_vanilla_guard_distr(
guard_to_bw, bw_resource)
rel_cost = relays.get_cost(bw_resource) / (
relays.get_cost(bw_resource) + sum(guard_to_cost.values()))
best_as_untargeted = best_as_dict[str(bw_resource)]
untargeted_prob_triple = untargeted_prob(client_as_lst,
[best_as_untargeted],
[bw_resource], pfi_instance)
print(bw_resource)
print("avg")
print(untargeted_prob_triple[0] / v_prob)
print(untargeted_prob_triple[0] / rel_cost)
print("max")
print(untargeted_prob_triple[2][1] / v_prob)
print(untargeted_prob_triple[2][1] / rel_cost)
def lastor_init_ratios(client_coord, fp_to_bw, fp_to_coord, fp_to_cost):
"""
Returns selection probability to cost ratio for client_coord
before applying redistribution algorithm to LASTor.
"""
total_cost = sum(fp_to_cost.values())
ratios = dict.fromkeys(fp_to_bw, 0)
distr = lastor.compute_lt_selection_probs(client_coord, fp_to_coord)
for fp, prob in distr.items():
cost = relays.get_cost(fp_to_bw[fp])
ratio = prob * total_cost / cost
ratios[fp] = ratio
return ratios
def denasa_init_ratios(client_as, fp_to_bw, fp_to_as, fp_to_cost):
"""
Returns selection probability to cost ratio for client_as
before applying redistribution algorithm to DeNASA.
"""
total_cost = sum(fp_to_cost.values())
ratios = dict.fromkeys(fp_to_bw, 0)
distr = denasa.compute_dn_selection_probs(client_as, fp_to_bw, fp_to_as,
all_ases, pfi_instance)
for fp, prob in distr.items():
cost = relays.get_cost(fp_to_bw[fp])
ratio = prob * total_cost / cost
ratios[fp] = ratio
return ratios
def disp_targeted_split_table(client_as, bw_resource, num_relays_lst):
"""
Prints table showing varying relays effect targeted attack.
"""
best_as = cr.compute_attack_as([client_as], client_to_all_res,
client_to_guard_res, guard_to_bw, ip_to_as,
sample_size)
v_guard_probs, v_prob = vanilla.compute_vanilla_guard_distr(
guard_to_bw, bw_resource)
for num_relays in num_relays_lst:
all_guard_probs, mal_prob = cr.compute_cr_guard_distr(
client_as, best_as, client_to_all_res, client_to_guard_res,
guard_to_bw, ip_to_as, num_relays, bw_resource, alpha, sample_size)
cost = num_relays * relays.get_cost(bw_resource / num_relays)
tot_cost = sum(guard_to_cost.values()) + cost
rel_cost = cost / tot_cost
print(
f"{num_relays} | relCost: {rel_cost} | VT prob: {v_prob} | targ CR prob: {mal_prob} | AS{best_as}"
)
def disp_untargeted_split_table_avg(client_as_lst, bw_resource,
num_relays_lst):
"""
Prints table showing varying relays effect on untargeted attack
(average success).
Format:
Relay # | relCost | vanilla prob | avg untarg prob | max untarg prob with client AS
"""
v_guard_probs, v_prob = vanilla.compute_vanilla_guard_distr(
guard_to_bw, bw_resource)
for num_relays in num_relays_lst:
inDir = f"untarg{num_relays}"
mal_guard_as_lst = []
for i in range(num_relays):
fname = f"{inDir}/tempbeststate{i}.json"
best_state = json.load(open(fname))
mal_guard_as_lst.append(best_state[1])
bw_per_guard = bw_resource / num_relays
mal_guard_bw_lst = [bw_per_guard] * num_relays
res_triple = untargeted_prob(client_as_lst, mal_guard_as_lst,
mal_guard_bw_lst, pfi_instance)
avg_prob = res_triple[0]
hi_prob = res_triple[2] # tuple of (client AS, prob)
cost = num_relays * relays.get_cost(bw_resource / num_relays)
tot_cost = sum(guard_to_cost.values()) + cost
rel_cost = cost / tot_cost
print(
f"{num_relays} | relCost: {rel_cost} | VT prob: {v_prob} | avg prob: {avg_prob} | max prob: {hi_prob}"
)
sys.path.append('..')
sys.path.append('../vanilla')
import numpy as np
import argparse
import lastor as lt
import vanilla
import relays
import json
import pickle
import operator
import geopy.distance
guard_to_bw = pickle.load(open("../guard_info/guard_to_bw.pickle", "rb"))
guard_to_cost = {guard: relays.get_cost(bw) for guard, bw in guard_to_bw.items()}
relay_ips = [ip.strip() for ip in open('../data/relay_ips.txt', 'r').readlines()]
def make_prob_matrix(client_lst, mal_coord, bw_resources_lst):
"""
Returns a numpy matrix where every row corresponds to the bandwidth
col_0: Vanilla Tor selection probability
col_1: Untargeted LT average selection probability (across all clients)
col_2: Untargeted LT lowest selection probability
col_3: Untargeted LT highest selection probability
client_lst: client coordinates
mal_coord: malicious guard placement (single relay)
bw_resources_lst: list of bandwidth resources
"""
ip_to_as = json.load(open("../guard_info/ip_to_as.json"))
all_ases = [
asn.strip() for asn in open("../data/relay_ases.txt", 'r').readlines()
]
guard_to_bw = pickle.load(open("../guard_info/guard_to_bw.pickle", "rb"))
fp_to_bw = {g.fingerprint: bw for (g, bw) in guard_to_bw.items()}
fp_to_as = {
g.fingerprint: ip_to_as[g.address]
for (g, bw) in guard_to_bw.items()
}
fp_to_coord = pickle.load(open("../guard_info/guard_fps_to_coord.pickle",
"rb"))
fp_to_cost = {fp: relays.get_cost(bw) for fp, bw in fp_to_bw.items()}
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument('--dump', action='store_true')
return parser.parse_args()
# --------------------------------------------------------------
def counterraptor_init_ratios(client_as, fp_to_bw, fp_to_as, fp_to_cost):
"""
Returns selection probability to cost ratio for client_as
before applying redistribution algorithm to Counter-RAPTOR.
def counterraptor_redistr(client_as, fp_to_bw, fp_to_as, threshold):
"""
Redistributes Counter-RAPTOR guard probabilities.
threshold = selection probability to fraction of the cost of the network
contributed
"""
if threshold < 1:
print("Error: threshold must be at least 1.")
return None
client_to_all_res = json.load(
open("../counterraptor/all_reachable_resilience.json"))
client_to_guard_res = json.load(
open("../counterraptor/cg_resilience.json"))
alpha = 0.5
sample_g = 0.1
total_cost = 0
for fp, bw in fp_to_bw.items():
total_cost += relays.get_cost(bw)
thresh = threshold / total_cost
ratios = dict.fromkeys(fp_to_bw, 0)
excess = 1
curr_fps = list(fp_to_bw.keys())
cnt = 1
while excess > 0:
print(f'----- Round {cnt} -----')
cnt += 1
if len(curr_fps) == 0:
break
curr_fp_to_bw = {fp: fp_to_bw[fp] for fp in curr_fps}
curr_fp_to_as = {fp: fp_to_as[fp] for fp in curr_fps}
curr_distr = counterraptor.compute_cr_selection_probs(
client_as, curr_fp_to_bw, curr_fp_to_as, client_to_all_res,
client_to_guard_res, alpha, sample_g * len(curr_fp_to_bw))
print(f'sum distr (~1): {sum(curr_distr.values())}')
norm = excess
print(f'norm: {norm}')
num_over = 0
excess = 0
for fp in list(curr_fps):
cost = relays.get_cost(fp_to_bw[fp])
ratio = curr_distr[fp] / cost * norm
ratios[fp] += ratio
if ratios[fp] > thresh:
excess += ((ratios[fp] - thresh) * cost)
ratios[fp] = thresh
curr_fps.remove(fp)
num_over += 1
print(f'num over thresh: {num_over}')
redistr = {}
for fp, val in ratios.items():
prob = val * relays.get_cost(fp_to_bw[fp])
redistr[fp] = prob
return redistr
import counterraptor as cr
import vanilla
import json
import pickle
import relays
# ------- init info ------------------
client_to_all_res = json.load(open("all_reachable_resilience.json"))
client_to_guard_res = json.load(open("cg_resilience.json"))
client_as_lst = list(client_to_guard_res.keys())
ip_to_as = json.load(open("../guard_info/ip_to_as.json"))
guard_to_bw = pickle.load(open("../guard_info/guard_to_bw.pickle", "rb"))
guard_to_cost = {
guard: relays.get_cost(bw)
for guard, bw in guard_to_bw.items()
}
alpha = 0.5
sample_size = int(0.1 * len(guard_to_bw)) # g = 0.1
# ------------------------------------
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument("--target_as")
return parser.parse_args()
