refactor analysis and is_around_message_interval

This commit is contained in:
Youngjoon Lee 2024-05-17 12:24:19 +09:00
parent fcf16153aa
commit 4655c34c71
No known key found for this signature in database
GPG Key ID: 09B750B5BD6F08A2
3 changed files with 63 additions and 42 deletions

52
mixnet/v2/sim/analysis.py Normal file
View File

@ -0,0 +1,52 @@
import pandas as pd
import seaborn
from matplotlib import pyplot as plt
from simulation import Simulation
class Analysis:
def __init__(self, sim: Simulation):
self.sim = sim
def run(self):
self.message_size_distribution()
self.messages_emitted_around_interval()
self.mixed_messages_per_node_over_time()
def message_size_distribution(self):
df = pd.DataFrame(self.sim.p2p.message_sizes, columns=["message_size"])
print(df.describe())
def messages_emitted_around_interval(self):
df = pd.DataFrame(
[(node.id, cnt, node.id < len(self.sim.config.real_message_prob_weights))
for node, cnt in self.sim.p2p.senders_around_interval.items()],
columns=["node_id", "msg_count", "expected"]
)
plt.figure(figsize=(10, 6))
seaborn.barplot(data=df, x="node_id", y="msg_count", hue="expected", palette={True: "red", False: "blue"})
plt.title("Messages emitted around the promised interval")
plt.xlabel("Sender Node ID")
plt.ylabel("Msg Count")
plt.legend(title="expected")
plt.show()
def mixed_messages_per_node_over_time(self):
dataframes = []
for mixed_msgs_per_node in self.sim.p2p.mixed_msgs_per_window:
df = pd.DataFrame([(node.id, cnt) for node, cnt in mixed_msgs_per_node.items()],
columns=["node_id", "msg_count"])
dataframes.append(df)
observation_times = range(len(dataframes))
df = pd.concat([df.assign(Time=time) for df, time in zip(dataframes, observation_times)], ignore_index=True)
df = df.pivot(index="Time", columns="node_id", values="msg_count")
plt.figure(figsize=(12, 6))
for column in df.columns:
plt.plot(df.index, df[column], marker="o", label=column)
plt.title("Mixed messages in each mix over time")
plt.xlabel("Time")
plt.ylabel("Msg Count")
plt.legend(title="Node ID")
plt.grid(True)
plt.show()

View File

@ -1,10 +1,7 @@
import argparse
import matplotlib.pyplot as plt
import pandas as pd
import seaborn
from config import Config
from mixnet.v2.sim.analysis import Analysis
from simulation import Simulation
if __name__ == "__main__":
@ -16,40 +13,6 @@ if __name__ == "__main__":
sim = Simulation(config)
sim.run()
# Stat the distribution of message sizes
df = pd.DataFrame(sim.p2p.message_sizes, columns=["message_size"])
print(df.describe())
# Visualize the nodes emitted messages around the promised interval
df = pd.DataFrame(
[(node.id, cnt, node.id < len(config.real_message_prob_weights))
for node, cnt in sim.p2p.senders_around_interval.items()],
columns=["NodeID", "MsgCount", "Expected"]
)
plt.figure(figsize=(10, 6))
seaborn.barplot(data=df, x="NodeID", y="MsgCount", hue="Expected", palette={True: "red", False: "blue"})
plt.title("Messages emitted around the promised interval")
plt.xlabel("Sender Node ID")
plt.ylabel("Msg Count")
plt.legend(title="Expected")
plt.show()
# Analyze the number of mixed messages per node per observation window
dataframes = []
for mixed_msgs_per_node in sim.p2p.mixed_msgs_per_window:
df = pd.DataFrame([(node.id, cnt) for node, cnt in mixed_msgs_per_node.items()], columns=["NodeID", "MsgCount"])
dataframes.append(df)
observation_times = range(len(dataframes))
df = pd.concat([df.assign(Time=time) for df, time in zip(dataframes, observation_times)], ignore_index=True)
df = df.pivot(index="Time", columns="NodeID", values="MsgCount")
plt.figure(figsize=(12, 6))
for column in df.columns:
plt.plot(df.index, df[column], marker='o', label=column)
plt.title('Mixed messages in each mix over time')
plt.xlabel('Time')
plt.ylabel('Msg Count')
plt.legend(title='Node ID')
plt.grid(True)
plt.show()
Analysis(sim).run()
print("Simulation complete!")

View File

@ -3,6 +3,7 @@ import random
from collections import defaultdict
import simpy
from simpy.core import SimTime
from config import Config
from sphinx import SphinxPacket
@ -30,11 +31,11 @@ class P2p:
# but we accept SphinxPacket as well because we don't implement Sphinx deserialization.
def broadcast(self, sender, msg: SphinxPacket | bytes):
self.log("Broadcasting a msg: %d bytes" % len(msg))
# Adversary
self.message_sizes.append(len(msg))
self.mixed_msgs_per_window[-1][sender] -= 1
now_frac, now_int = math.modf(self.env.now)
if now_int % self.config.message_interval == 0 and now_frac <= self.config.max_message_prep_time:
if self.is_around_message_interval(self.env.now):
self.senders_around_interval[sender] += 1
# Yield 0 to ensure that the broadcast is done in the same time step.
@ -52,6 +53,11 @@ class P2p:
self.mixed_msgs_per_window[-1][node] += 1
self.env.process(node.receive_message(msg))
# TODO: Move to a separate class `Adversary`.
def is_around_message_interval(self, time: SimTime):
now_frac, now_int = math.modf(time)
return now_int % self.config.message_interval == 0 and now_frac <= self.config.max_message_prep_time
# TODO: Move to a separate class `Adversary`.
def update_observation_window(self):
while True: