Professional Documents
Culture Documents
ICATM Paper Template
ICATM Paper Template
ICATM Paper Template
The survey paper of two phases: training and testing. In the 2.2.1 Dataset size and diversity
training phase, the attacker collects a set of website traces and The performance of the model is highly dependent on the
extracts their corresponding features. The features are then dataset and diversity of the dataset used to train it. A larger and
clustered using homology analysis to generate a set of more diverse dataset is likely to lead to better performance.
representative templates. In the testing phase, the attacker
2.2.2 Graph structure
captures the traffic of the target user and extracts the features The structure of the graph used to represent the network traffic
of each packet. The attacker then compares the extracted can have a significant impact on the performance of the model.
features with the representative templates to identify the Different graph structures can capture different aspects of the
visited web pages. It does not rely on the assumption that the network traffic, so it's important to choose an appropriate
attacker has prior knowledge of the website's content. It has a structure.
high accuracy rate, which means that it is effective in 2.2.3 Attention mechanism
identifying the websites visited by a user. The attention mechanism used in the GAPN can affect the
performance of the model. Different attention mechanisms can
2. WEBSITE FINGERPRINTING focus on different parts of the input data and capture different
SYSTEM patterns.
2.3 Proposed System Architecture graph (DAG) that captures the temporal ordering of the
SSL/TLS packets.
for the FPR and vice-versa effectively, allowing us to tackle the sequences accurately from the request-response structure, with
base rate fallacy. We have seen that our new WLLCC algorithm the network conditions as input, we could significantly improve
for Wa-kNN has lifted the competitiveness of the simple and website fingerprinting. Data collection would be much faster,
fast k-NN classifier to that of SVMs for website fingerprinting. as we would not have to wait for Tor’s latency to load web
The algorithm may be useful for k-NN classification in general, pages. We would also be able to build better classifiers that
and it can also be used in clustering. ignore random noise such as advertisements.