belarus-2020-2021/scripts/evaluate-blockages.py

#!/usr/bin/env python3

import csv
import os

# Algorithm used in https://research.torproject.org/techreports/blocking-2011-09-15.pdf
def evaluate_loesing (fingerprint):
    fingerprint = fingerprint.upper()
    bridge_ips_max = None

    # First get the max value in whole range
    with open (f"data/bridge_data_cleaned/{fingerprint}", 'r') as file:
        bridge_data = csv.reader(file, delimiter=',')

        for row in bridge_data:
            bridge_ips_today = int(row[1])

            if bridge_ips_max is None:
                if bridge_ips_today > 0:
                    bridge_ips_max = bridge_ips_today
                    #print (f"{fingerprint} bridge_ips_max set to {bridge_ips_today} on {row[0]}")
            elif bridge_ips_today > bridge_ips_max:
                bridge_ips_max = bridge_ips_today
                #print (f"{fingerprint} bridge_ips_max set to {bridge_ips_today} on {row[0]}")

    # Now evaluate
    with open (f"data/bridge_data_cleaned/{fingerprint}", 'r') as file:
        bridge_data = csv.reader(file, delimiter=',')

        for row in bridge_data:
            bridge_ips_today = int(row[1])

            if not bridge_ips_max is None and bridge_ips_max >= 100 and bridge_ips_today <= 32:
                return row[0]

    return None

# Consider a bridge blocked if it receives <= t connections in one day,
# when it has previously received >= m connections. Note that t=32,
# m=100 is NOT the same as Loesing's algorithm, as we consider m only
# for previous days, while Loesing's algorithm considers all days within
# the period. t must be less than m.
def is_blocked_abs (t, m, bridge_ips_max, bridge_ips_today):
    if bridge_ips_max is None:
        return False

    return bridge_ips_max >= m and bridge_ips_today <= t

def evaluate_abs (t, m, fingerprint):
    fingerprint = fingerprint.upper()
    bridge_ips_max = None

    with open (f"data/bridge_data_cleaned/{fingerprint}", 'r') as file:
        bridge_data = csv.reader(file, delimiter=',')

        for row in bridge_data:
            bridge_ips_today = int(row[1])

            if not bridge_ips_max is None:
                if is_blocked_abs (t, m, bridge_ips_max, bridge_ips_today):
                    return row[0]

                if bridge_ips_today > bridge_ips_max:
                    bridge_ips_max = bridge_ips_today

            # Start bridge_ips_max only when we have a non-zero
            # connection count
            elif bridge_ips_today > 0:
                bridge_ips_max = bridge_ips_today

        # If we got here, the bridge is not blocked
        return None

# Consider a bridge blocked if we observe a drop of at least d from the
# previous day's count. d must be at least 8.
def is_blocked_rel (d, bridge_ips_yesterday, bridge_ips_today):
    if bridge_ips_yesterday is None:
        return False

    return bridge_ips_today <= bridge_ips_yesterday - d

def evaluate_rel (d, fingerprint):
    fingerprint = fingerprint.upper()
    bridge_ips_yesterday = None
    bridge_ips_today = None

    with open (f"data/bridge_data_cleaned/{fingerprint}", 'r') as file:
        bridge_data = csv.reader(file, delimiter=',')

        for row in bridge_data:
            bridge_ips_today = int(row[1])

            if is_blocked_rel (d, bridge_ips_yesterday, bridge_ips_today):
                return row[0]

            bridge_ips_yesterday = bridge_ips_today

# Remove any previous blocked_* files, start over
if os.path.exists (f"data/blocked_loesing"):
    os.remove (f"data/blocked_loesing")

for t in range (8, 40, 8):
    for m in range (t+8, 112, 8):
        if os.path.exists (f"data/blocked_abs_{t}_{m}"):
            os.remove (f"data/blocked_abs_{t}_{m}")

for d in range (8, 112, 8):
    if os.path.exists (f"data/blocked_rel_{d}"):
        os.remove (f"data/blocked_rel_{d}")

with open ("data/all-bridges", 'r') as all_bridges:
    for fingerprint in all_bridges:
        fingerprint = fingerprint.strip()
        if fingerprint:
            # Evaluate with Loesing's algorithm
            blocked = evaluate_loesing (fingerprint)
            if not blocked is None:
                with open (f"data/blocked_loesing", 'a') as f:
                    f.write(f"{fingerprint},{blocked}\n")

            # Absolute threshold with t and m
            for t in range (8, 40, 8):
                for m in range (t+8, 112, 8):
                    blocked = evaluate_abs (t, m, fingerprint)
                    if not blocked is None:
                        with open (f"data/blocked_abs_{t}_{m}", 'a') as f:
                            f.write (f"{fingerprint},{blocked}\n")

            # Relative threshold with d
            for d in range (8, 112, 8):
                blocked = evaluate_rel (d, fingerprint)
                if not blocked is None:
                    with open (f"data/blocked_rel_{d}", 'a') as f:
                        f.write (f"{fingerprint},{blocked}\n")