📌 Concerning Repetitions in 24-Word Mnemonic Phrases
Self-taught | Specialized in Blockchain Security via hands-on testing and continuous research
👤 By GUIAR OQBA – ELKANTARA
⚠️ Note: This article contains a summarized version of the full technical report. To view the complete version including detailed images, visit:
👉 https://zenodo.org/records/15786076
📊 General Introduction & Tools Used
This is the third installment in a series analyzing entropy derived from 24-word mnemonic phrases as per the BIP-39 standard. This time, the focus is binary-level analysis, bypassing linguistic and purely statistical considerations.
🛠️ A custom C-based analysis tool was built for full control, avoiding reliance on high-level libraries.
🔧 Core Steps:
Convert 24-word phrases to 256-bit entropy.
Detect suspicious byte patterns:
0x00,0x55,0xAA,0xFF.Analyze bit counts, transitions, and frequency of specific bytes/nibbles.
Code Excerpt:
[Here you can embed a code block if desired. Omit if no snippet.]
⚙️ Section 2: Large-Scale Execution & Binary Analysis
The tool was run on over 87,000 valid phrases. A refined subset of 16,650 was selected due to suspected anomalies.
🧬 1. Critical Byte Repetition
The following bytes were found repeatedly:
0x00– null or padding0x55– binary 010101010xAA– binary 101010100xFF– full-value bytes
📈 Counts:
Phrases with
0x00,0x55,0xAA: 4619 eachPhrases with
0xFF: Less frequent but present
🖼️ [Image Placeholder 1: Terminal screenshot showing grep -c command]
🧩 2. Dual Byte Patterns in Same Phrase
Many phrases contained multiple critical bytes in the same entropy block — a major red flag for non-randomness.
🖼️ [Image Placeholder 2: grep -E output showing multi-byte matches]
📈 3. Bit Count & Transition Analysis
The script calculated:
1s vs. 0s count
Bit transitions (0→1 or 1→0)
🧠 Some phrases had:
Perfectly balanced
128/128bitsSuspiciously low or high transition counts (e.g., 104 or 142)
🖼️ [Image Placeholder 3: Bit transition results]
🧪 4. Nibble-Level Pattern Repetition
Even at the 4-bit (nibble) level, recurring values like 0x5, 0xA, 0x0 were observed.
🖼️ [Image Placeholder 4: Nibble frequency plot]
📂 5. About suspicious_seeds.txt
This file logs all phrases with:
Critical byte presence
Unusual bit balance
Extreme transitions
Repetitive patterns at byte/nibble level
🖼️ [Image Placeholder 5: Sample from suspicious_seeds.txt with line numbers]
🧠 Technical Interpretation
🔹 1. Byte Repetition: Coincidence or Cause?
Having 0x00, 0x55, or 0xAA in ~28% of phrases is statistically implausible in truly random entropy.
🔍 Possible causes:
Template-based generators
Deterministic wallets
Encoding bugs
🔸 2. Triple Pattern Correlation
These bytes represent unique binary patterns:
| Byte | Binary |
| 0x55 | 01010101 |
| 0xAA | 10101010 |
| 0x00 | 00000000 |
Their co-occurrence indicates:
Debug-mode generation
Weak PRNGs or test data leakage
⚖️ 3. Bit Transition Extremes
Standard entropy shows ~127 transitions (±7). Phrases with 104 or 142 transitions are outside the secure envelope.
🏭 4. Industrial Patterns in Nibbles
These suggest:
Aesthetic padding or deterministic generation
Reuse across wallets or same app behavior
🧱 5. Template-Based or Cloning Behavior
Detected phrases with near-identical binary patterns suggest:
Partial reuse of entropy
Wallets created in structured batches
🔐 6. 12-word Phrases: No Such Issues
No such anomalies were found in 12-word phrases.
❓ Raises the question: Are 24-word generators more likely to leak structure?
🧾 Security Risk Assessment
⚠️ 1. Repeating Critical Bytes
Suggest weak entropy or deterministic sources
Brute-force becomes more feasible due to reduced key space
❗ 2. Identical Distributions
Repeated 1s/0s and nibbles violate randomness principles, hinting at systematic flaws.
🚨 3. BIP-39 Security Model Violation
Findings suggest:
Entropy is not truly random
Some generators behave deterministically
🌐 4. Implications for Web3 Wallets
Phrases may be recoverable via pattern-targeted attacks
Especially risky on Ethereum, Solana, etc.
✅ Final Thoughts
🧠 BIP-39 itself is not flawed. But:
Many generators produce weak or structured entropy
Security researchers should audit online or closed-source tools
📌 Takeaways
0x00,0x55,0xAApatterns aren’t random.Overlapping binary structures = cloning or templating.
Entropy space is shrinking = wallets at risk.
Urgent need for forensic studies into wallet generators.
👤 About the Author
GUIAR OQBA – ELKANTARA
Security Researcher | Blockchain Forensics
📧 Email: techokba@gmail.com
📬 Telegram: @Okba_elkantara
