diff --git a/images/transaction-analysis/.DS_Store b/images/transaction-analysis/.DS_Store
new file mode 100644
index 0000000..7a4283c
Binary files /dev/null and b/images/transaction-analysis/.DS_Store differ
diff --git a/images/transaction-analysis/100-blocks-tx-size.png b/images/transaction-analysis/100-blocks-tx-size.png
new file mode 100644
index 0000000..c1a5fbe
Binary files /dev/null and b/images/transaction-analysis/100-blocks-tx-size.png differ
diff --git a/images/transaction-analysis/length.png b/images/transaction-analysis/length.png
new file mode 100644
index 0000000..8ac71ce
Binary files /dev/null and b/images/transaction-analysis/length.png differ
diff --git a/images/transaction-analysis/mempool-tx-size.png b/images/transaction-analysis/mempool-tx-size.png
new file mode 100644
index 0000000..08803d8
Binary files /dev/null and b/images/transaction-analysis/mempool-tx-size.png differ
diff --git a/traffic-analysis.md b/traffic-analysis.md
new file mode 100644
index 0000000..0b2e248
--- /dev/null
+++ b/traffic-analysis.md
@@ -0,0 +1,53 @@
+# I still see you!
+
+## Problem
+
+Sending transactions across the P2P network is sensitive data.
+
+![show transaction tracking based on size](./images/transaction-analysis/length.png)
+
+Global passive observers can differentiate whether it's this tx or that tx from the size of the packet!
+
+## Statistics
+
+### Common transaction size (in bytes) in the last 100 blocks
+![show common transaction size in last 100 blocks](./images/transaction-analysis/100-blocks-tx-size.png)
+- 96.7 % of transactions below 1000 bytes
+- 99% of transactions below 3000 bytes
+
+### Common transaction size (in bytes) in the mempool
+![show common transaction size in mempool](./images/transaction-analysis/mempool-tx-size.png)
+- 89.6 % of transactions below 1000 bytes
+- 92% of transactions below 3000 bytes
+
+## Solution
+
+1. Pad transaction messages
+
+- last 100 blocks
+
+  |                          | pad to a fixed size=1000 | pad to a fixed size=3000 | pad to a dynamic size |
+  |--------------------------|--------------------------|--------------------------|-----------------------|
+  | how much padding?        | 1000                     | 3000                     | Padmé algo            |
+  | what is extra bandwidth? | 43 MB                    | 162 MB                   | 10.9 GB               |
+
+- mempool
+
+   |                          | pad to a fixed size=1000 | pad to a fixed size=3000 | pad to a dynamic size |
+   |--------------------------|--------------------------|--------------------------|-----------------------|
+   | how much padding?        | 1000                     | 3000                     | Padmé algo            |
+   | what is extra bandwidth? | 37 MB                    | 139.8 MB                 | 123 MB                |
+
+2. Broadcast decoy messages with size (transaction message)
+
+## Code
+
+- branch - https://github.com/stratospher/bitcoin/tree/2024-05-tx-traffic
+- jupyter notebook - https://colab.research.google.com/drive/1q2fLkxuAiqr2hlAEQ4r6x2Lx9bTh4bN0?usp=sharing
+
+## Conclusion
+
+|                 | advantage                | disadvantage                                                           |
+|-----------------|--------------------------|------------------------------------------------------------------------|
+| fixed padding   | bandwidth efficient      | can't hide the outliers + optimal padding might vary based on data set |
+| dynamic padding | covers even the outliers | too much bandwidth                                                     |