Pith Number

pith:E4LQU2RR

pith:2026:E4LQU2RRAXUKRY2XPNP4RWKYKQ

not attested not anchored not stored refs resolved

Backdoor Channels Hidden in Latent Space: Cryptographic Undetectability in Modern Neural Networks

Eirik Reiestad, Inga Str\"umke, Kristian Gj{\o}steen, Marte Eggen

Neural networks can hide backdoors as statistically indistinguishable latent directions, reducing detection to an intractable hypothesis test on model parameters.

arxiv:2605.13214 v1 · 2026-05-13 · cs.CR · cs.LG

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{E4LQU2RRAXUKRY2XPNP4RWKYKQ}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

if exploitable channels within a network's latent space are statistically indistinguishable from naturally learned directions, an attacker need not introduce foreign structure but can instead exploit the geometry the network already possesses.

C2weakest assumption

The hypothesis test between clean and backdoored parameter distributions is intractable in practice for state-of-the-art models; this is stated as a conjecture without a formal reduction or hardness proof.

C3one line summary

Backdoors can be realized as statistically natural latent directions in modern neural networks, achieving high attack success with negligible clean accuracy loss and resisting existing defenses.

References

31 extracted · 31 resolved · 5 Pith anchors

[1] Backdoor attacks and defenses in computer vision domain: A survey.arXiv preprint arXiv:2509.07504, 2025 2025

[2] Complexity theoretic lower bounds for sparse principal component detection 2013

[3] Computational Lower Bounds for Sparse PCA 2013 · arXiv:1304.0828

[4] Brennan and Guy Bresler 2019

[5] Data free backdoor attacks.Advances in Neural Information Processing Systems, 37:23881–23911, 2024 2024

Receipt and verification

First computed	2026-05-18T03:08:48.498585Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

27170a6a3105e8a8e3577b5fc8d95854151c879d172e04ba07e0a4ed0b917ed2

Aliases

arxiv: 2605.13214 · arxiv_version: 2605.13214v1 · doi: 10.48550/arxiv.2605.13214 · pith_short_12: E4LQU2RRAXUK · pith_short_16: E4LQU2RRAXUKRY2X · pith_short_8: E4LQU2RR

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/E4LQU2RRAXUKRY2XPNP4RWKYKQ \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: 27170a6a3105e8a8e3577b5fc8d95854151c879d172e04ba07e0a4ed0b917ed2

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "ff1d52e33a35d77f68cff34a954b3ac8f8038e05c1445207f5618451866f2d88",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CR",
    "submitted_at": "2026-05-13T09:06:25Z",
    "title_canon_sha256": "e7b8335d6cd6ec80eabb3f99a2153738a0a57aaa2990561a545ac320a9df870f"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.13214",
    "kind": "arxiv",
    "version": 1
  }
}