Project Power

Technical Walkthrough

Writeup

Challenge

• Name: Project-Power
• Category: Hardware
• Difficulty: Medium
• Mode: hybrid

Summary

Project Power is a remote hardware side-channel challenge. The provided interface allows chosen 16-byte plaintext encryption on a standard AES-128 device and returns a captured power trace. A second option verifies a submitted AES key and returns the flag if the key is correct.

The solution was to collect aligned power traces for random chosen plaintexts, run correlation power analysis against the first AES round S-box Hamming-weight leakage model, recover the 16-byte AES key, and submit it to the verifier. Raw key material and the flag are kept in loot/.

Artifact Inventory

The original ZIP is preserved at files/a12c737e-f8bd-4499-a3e1-14de70ffc81a.zip. Extraction produced:

• analysis/extracted/socket_interface.py: sample socket client with option 1 for trace collection and option 2 for key verification.
• analysis/extracted/remote_lab_layout.png: lab diagram confirming standard AES-128 encryption and oscilloscope power-trace capture.

The extracted file types and hashes are recorded in analysis/file-types.txt and analysis/sha256sums.txt.

Analysis

socket_interface.py documents the protocol: send option 1 with a 16-byte plaintext to receive base64-encoded NumPy trace data, or send option 2 with a hex-encoded AES key to verify it. The lab diagram confirms the device encrypts arbitrary 16-byte plaintexts with standard AES-128.

The first probe in analysis/trace-probe-1.txt validated the live response format. Each trace decoded as 1042 float64 samples. Since plaintext is chosen and AES is standard, the cheapest key-recovery path is first-round CPA:

1. For each trace and key-byte guess, compute SBOX[plaintext_byte ^ guess].
2. Convert that hypothetical intermediate to Hamming weight.
3. Correlate the hypothetical leakage vector with every trace sample.
4. Select the byte guess with the highest absolute correlation.

The final run used 250 traces saved in analysis/traces/project_power_250.npz. The redacted CPA summary in analysis/cpa-summary-redacted.txt shows all 16 bytes recovered with strong score separation. Raw recovered key material was moved to loot/.

Solve

The reproducible solver is solve/solve.py. It can collect traces, run CPA, write the recovered key to loot/, and submit the key for verification.

Example:

cd <local workspace>
python3 solve/solve.py --host <TARGET> --port 32006 --count 250 --delay 0.03 --force-collect

For the completed run, the saved traces can be reused without recollecting:

python3 solve/solve.py --traces analysis/traces/project_power_250.npz --no-verify

The successful key-check response was saved to loot/flag-candidate.txt and captured by the harness into loot/flag.txt.

Flag

Raw flag is stored in loot/flag.txt and intentionally not reproduced here.

Lessons

• For AES power traces with chosen plaintext, first-round S-box Hamming-weight CPA is the right first validation before trying more complex alignment or later-round models.
• Do not wait for EOF if a remote lab socket keeps connections open; stop reading once the base64 trace decodes to a valid float64 array.
• Treat recovered keys like challenge secrets: keep raw key and flag material in loot/, and use redacted analysis summaries for documentation.

Source-Backed Dossier

The sections below are merged from companion Markdown notes for the same case. They are rendered after sanitization so the article stays precise without publishing raw flags, credentials, or target-specific secrets.

Notes

Scope

• Challenge: Project-Power
• Category: Hardware
• Difficulty: Medium
• Mode: hybrid
• Remote instance: <TARGET>:32006
• Start time: 2026-06-13T11:06:19Z
• Operator: harness
• State file: challenge-state.json

Harness Status

• Current phase: see challenge-state.json
• Next allowed actions: see next-action.json
• Raw flags and sensitive material stay in loot/ only. Do not paste them here.

Artifact Inventory

Evidence Ledger

Key Findings

• The lab exposes a chosen-plaintext AES-128 encryption oracle that returns power traces, plus a separate key verification option.
• Power traces decode as fixed-length float64 arrays with 1042 samples.
• First-round AES S-box Hamming-weight CPA over chosen plaintexts was sufficient; 250 traces gave clear per-byte correlation margins.
• Raw recovered AES key material and raw flag responses are stored under loot/ only.

RAG / Advisory Memory

RAG output is advisory only. Record evaluated retrievals with:

scripts/challenge_harness.py rag-record <workspace> --query "..." --tag MATCHED|PARTIAL|MISSING|<secret redacted>|GENERIC --validation "..."

Secrets/Flags

Raw flags and sensitive material stay in loot/ only. Use scripts/challenge_harness.py capture-flag to validate and record flag capture without printing the value.

Memory Summary

Metadata

• Platform: HackTheBox Challenges
• Category: Hardware
• Challenge: Project-Power
• Difficulty: Medium
• Source workspace: <local workspace>

Validated Solve Chain

Concepts only. Do not include raw flags, reusable credentials, tokens, cookies, private keys, or live secrets.

1. Provided hardware challenge exposed a socket interface for chosen-plaintext AES trace collection and a separate key verification endpoint.
2. Live trace probe confirmed base64-encoded NumPy float64 power traces with fixed sample length.
3. Collected a moderate batch of random chosen plaintext traces.
4. Recovered the AES-128 key with first-round SubBytes Hamming-weight correlation power analysis.
5. Submitted the recovered key to the remote verifier and captured the flag.

Reusable Lessons

• For chosen-plaintext AES power traces, first-round S-box Hamming-weight CPA should be the first validation path.
• Socket trace readers should stop on valid decoded trace data instead of assuming the service closes the connection.
• Store raw recovered keys and verifier responses in loot/; keep analysis outputs redacted.

Dead Ends

• Private CTF RAG had no useful prior memory for this challenge; direct artifacts and local CPA validation were sufficient.
• A 20-trace smoke test validated the implementation but was not enough for a reliable final key.

Tool Quirks

• binwalk was unavailable locally, but the archive contained only a PNG diagram and Python socket client, so firmware carving was not required.
• The socket service could be slow per trace; 250 traces took several minutes but produced stable CPA margins.

Evidence Paths

• analysis/extracted/socket_interface.py
• analysis/extracted/remote_lab_layout.png
• analysis/trace-probe-1.txt
• analysis/traces/project_power_250.npz
• analysis/cpa-summary-redacted.txt
• solve/solve.py
• loot/aes-key.txt
• loot/flag.txt

Ingestion Decision

• Proposed for LightRAG: yes
• Requires user approval before ingestion: yes

Hypothesis Board

Keep no more than 3 active hypotheses on Easy/Medium and 5 on Hard unless the user explicitly asks for breadth.

Closed Branches