Proposal · Ethereum Trillion Dollar Security (1TS) Initiative

Security under observation.

LeakDetect AI is an open-source framework that detects, measures, and simulates hidden information leakage across Ethereum wallets, AI agents, smart contracts, and multi-chain systems.

leakdetect ~ observability-trace · wallet 0xA1c…7e3
live
Vulnerability Score
0.78
high
Key Recovery Probability
0.42
warn
Information Leakage Rate
1.6 b/s
warn
capture timing gas signing rpc xchain
simulate behavioral_inference on 12,488 traces
infer side_channel via deep-learning ensemble
! leakage detected: signing-time variance correlates with nonce parity (p < 0.001)
TIMING_VARIANCEGAS_PATTERNNONCE_PARITYRPC_FINGERPRINTSIGN_CADENCEBRIDGE_RELAYAGENT_TRACEMETADATA_DRIFTKEY_RECOVERY_PROBSTEALTH_LEAK_INDEXTIMING_VARIANCEGAS_PATTERNNONCE_PARITYRPC_FINGERPRINTSIGN_CADENCEBRIDGE_RELAYAGENT_TRACEMETADATA_DRIFTKEY_RECOVERY_PROBSTEALTH_LEAK_INDEX
// paradigm shift
Instead of asking "can this system execute safely?" LeakDetect asks: what does this system unintentionally reveal while operating?
01: the problem

The threats no one is watching.

Ethereum security today focuses on visible failures: contract exploits, phishing, wallet compromise. But systems can function correctly while unintentionally leaking sensitive information.

Adversaries with deep learning, statistical inference, and side-channel analysis are turning observation itself into the attack surface.

01
Transaction timing
Microsecond variance reveals nonce entropy and signer state.
02
Gas usage patterns
Branch behavior in contracts becomes observable from receipts.
03
Signing behavior
HSM and wallet firmware leak through approval cadence.
04
Cross-chain execution traces
Relayers expose coordination metadata across L1/L2/L3.
05
API interaction patterns
RPC fingerprints uniquely identify wallets and agents.
06
Wallet metadata
Approvals, allowances, and dust create persistent fingerprints.
07
Autonomous agent coordination
Agent-to-agent flows leak strategy and intent.
08
Hardware side channels
Power, timing, and EM emissions remain underexamined on-chain.
02: alignment with 1ts

Mapped to Ethereum's trillion-dollar security themes.

ethereum.org/trillion-dollar-security ↗
UX & Blind Signingaligned
  • Signing-time leakage analysis
  • Wallet interaction observability
  • Approval-pattern fingerprinting
  • Autonomous signing behavior
Privacy & Metadata Exposurealigned
  • Behavioral deanonymization
  • Multi-chain traceability
  • Wallet fingerprinting
  • AI-agent operational inference
Cross-chain & Infrastructurealigned
  • Bridge interaction leakage
  • RPC metadata exposure
  • Cross-chain timing analysis
  • Infrastructure observability risk
AI-Generated Security Risksaligned
  • AI transaction agents
  • Autonomous execution frameworks
  • Agent-to-agent coordination
  • AI-generated contract flows
03: technical architecture

Five layers, one observability pipeline.

01
Signal Capture
Transaction timing, gas, signing behavior, wallet traces, network metadata, cross-chain relay timing.
02
Adversarial Simulation
Behavioral inference, timing correlation, cross-chain observability, AI-agent tracking attacks.
03
ML Inference Engine
Deep learning, NAS, AutoML, sequence analysis, time-series security modeling.
04
Leakage Quantification
Information leakage, key recovery probability, observability risk, operational fingerprinting.
05
Reporting & Visualization
Leakage dashboards, wallet risk analysis, agent observability maps, cross-chain reports.
04: quantification

Measurable security, not vibes.

LeakDetect introduces standardized metrics so that observability risk can be tracked, benchmarked, and reduced across releases: the way performance and coverage already are.

VS
0.00 – 1.00
Vulnerability Score
Overall exploitability estimate based on observable surface area.
KRP
0% – 100%
Key Recovery Probability
Likelihood of reconstructing secrets from leaked side-channels.
ILR
b/s
Information Leakage Rate
Amount of leaked information per unit of time, in bits/second.
SLI
index
Stealth Leakage Index
Difficulty of detecting the leakage by defenders and monitors.
05: research foundations

Eight years of leakage research, ported to Ethereum.

3y+
Open-source security research at the Research Center Trustworthy Data Science and Security, Ruhr-Universität Bochum.
5y+
Foundational AI and cryptographic security research at the Software Innovation Campus Paderborn (SICP).
Extends AutoSCA (ACM AISec, "Automated Detection of Side Channels in Cryptographic Protocols: DROWN the ROBOTs!") into wallets, AI agents, and multi-chain systems.
06: research results

Benchmarked leakage detection, measured against baselines.

Results from the LeakDetect AI research pipeline: AutoML based leakage detection on timing and error-code vulnerabilities, side-channel key recovery benchmarks, and validation against an industrial-grade TLS testing tool.

detection accuracy: timing based vulnerability
Technique10/9030/70Balanced
AutoML Mid-Point0.500.500.78
AutoML Log-Loss0.500.660.74
DL-LA Baseline0.460.490.50
AutoML estimators stay informative under class imbalance where the DL-LA baseline collapses toward random guessing.
detection accuracy: error-code based vulnerability
Technique10/9030/70Balanced
AutoML Mid-Point0.340.360.88
AutoML Log-Loss0.880.910.93
AutoML Cal Log-Loss0.870.910.94
AutoML Fischer's Exact Test0.820.620.95
AutoML Paired t-Test0.610.590.96
GMM Baseline0.330.330.33
MINE Baseline0.630.640.67
PC-Softmax Baseline0.690.690.70
DL-LA Baseline0.510.560.72
side-channel key recovery success rate (ARES 2022)
AES_RD
52% : 100%
AES_HD
38% : 60%
CHES CTF
2% : 2%
DPAv4
75% : 100%
ASCAD_f
46% : 100%
ASCAD_r
56% : 71%
ASCAD_f desync50
8% : 57%
ASCAD_r desync50
13% : 85%
ASCAD_f desync100
7% : 70%
ASCAD_r desync100
13% : 71%
All parameter combinations compared with 1D CNNs found by random search, across 4 systems with countermeasures and 6 without. Success rate is the share of attacks reaching guessing entropy of 1.
openssl tls evaluation
5
Vulnerable systems
3
Non-vulnerable systems
Validated by mapping the LeakDetect approach onto the industrial AutoSCA test tool from Achelos GmbH.
methodology
A model-agnostic leakage assessment score quantifies information leakage. AutoML based estimators provide a black-box automated detection path, scale to high-dimensional datasets, and remain precise on imbalanced data. Statistical gating uses one-sample t-test, Fischer's exact test and paired t-test with Holm-Bonferroni correction before a system is labelled vulnerable.
07: giveth update

LeakDetect: protecting Ethereum from hidden leakages.

We are grateful to everyone who supported LeakDetect through the Ethereum Security QF round. The project received direct donations plus ETH from the matching pool, with special thanks to The DAO Fund, Wintermute and all individual contributors.

01
Advanced the leakage-detection methodology
An automated, black-box approach to detect, measure, reproduce, understand and mitigate hidden information leakage, including timing, metadata, network behaviour, communication patterns and other side channels.
02
Expanded from benchmarks to real-world ecosystems
Collecting and preparing datasets across Ethereum, Zcash, Tornado Cash, wallet security, P2P/libp2p and cross-chain workflows to validate whether the methodology generalizes across environments.
03
Started ETH and FIL security research
Investigating Ethereum to Filecoin swap workflows for potential leakage across transaction flows, interoperability, networking and observable system behaviour.
04
Extended research toward P2P/libp2p security
Moving beyond application-level signals toward network metadata, timing, peer interactions, routing and protocol behaviour, connecting directly with our libp2p, IPFS, IPLD and multiformats work.
05
Increased ecosystem engagement
Presented two Internet-Drafts and two research papers at IETF 126 in Vienna, recommended for an Ethereum 1TS proposal, with continued discussions across the Ethereum security ecosystem.
06
Broadened the research vision
The LeakDetect AI plus IPFS/libp2p deck is now public and used as the shared reference for collaborators.
next milestones and use of QF funds
  • Complete real-world dataset collection and preparation.
  • Run automated leakage detection across Ethereum, Zcash, Tornado Cash, wallets and ETH-FIL workflows.
  • Extend analysis to P2P/libp2p network-level leakage.
  • Publish technical reports, datasets where appropriate, case studies and IETF materials.
  • Document reproducible findings, leakage scores and practical mitigations.
  • Refine milestones with Ethereum security and P2P contributors and continue the 1TS proposal.
  • Explore a Trillion Dollar Security Research Retreat bringing researchers, protocol developers and security engineers together.
Our broader goal is simple: identify what decentralized systems unintentionally reveal, and develop practical, reproducible ways to detect and mitigate that leakage before it becomes a security or privacy problem.
07: decentralized ai infrastructure

Scalable, private, and decentralized by default.

As LeakDetect AI scales across wallets, agents, and chains, we are building the infrastructure to keep security intelligence open, privacy-preserving, and censorship-resistant: from storage to inference.

IPFSprivacy-preserving storage

Audit logs, inference traces, and leakage datasets are stored on IPFS with content-addressing and encryption. No single point of failure. No custodial data silos.

libp2pdecentralized networking

P2P gossipsub and DHT routing enable AI agents to share threat signals, model updates, and observability metadata without centralized relays. Resilient by design.

Nillionzk training & inference

Privacy-preserving computation for ML training and inference. Secrets are split into shares across a decentralized network: no node ever sees the full model or raw data.

Hugging Faceopen model ecosystem

Foundation models, tokenizers, and evaluation pipelines integrated natively. Fine-tune leakage-detection transformers and publish them back to the community.

zk pipeline overview
01
Capture
Signals from wallets, agents, chains
02
Encrypt & Shard
Nillion secret sharing
03
zk Train / Infer
Private ML on distributed shares
04
Publish
IPFS + libp2p propagation
08: what we plan to build

Open infrastructure for observability-aware security.

01
Production-ready toolkit
  • CLI tooling
  • SDKs
  • Wallet integrations
  • AI-agent integrations
  • Multi-chain test infra
02
Public benchmarks & datasets
  • Standardized leakage datasets
  • Open benchmark suites
  • Reproducible pipelines
  • AI-agent observability corpora
03
Ecosystem integrations
  • Wallet providers
  • AI-agent frameworks
  • Infrastructure providers
  • Security monitoring
  • Multi-chain dApps
04
Security standardization
  • Leakage-aware metrics
  • Wallet observability standards
  • Agent operational assessment
  • Cross-chain metadata risk
05
Decentralized AI infrastructure
  • IPFS for privacy-preserving log storage
  • libp2p for decentralized AI networking
  • Nillion zk training & inference pipeline
  • Hugging Face model integration & fine-tuning
09: partnerships & field testing

Collaborations powering observability-aware security in the wild.

Achelos / AutoSCA
Side-channel analysis tooling
www.achelos.de/en/
Aave
Enterprise AI agent observability
Aspiring INVESTMENTS CORP
Investment & venture partner
www.aspiringapps.com
Agent0 (ERC-8004)
ERC-8004: Trustless Agents. Discover agents and establish trust through reputation and validation
Morpho
Health-AI observability pilots
Thales Security team led by Vibhor Bijoy
Security research & advisory
Samagra
Public-sector deployments
samagragovernance.in/
TickerValue
Partner platform & market intelligence
tickervalue.com
Park Pro Token
Field testing partner & token deployment
park-token-deploy.vercel.app/
libp2p
Decentralized peer-to-peer networking for AI agents
libp2p.io/
IPFS
Privacy-preserving, content-addressed log storage
ipfs.tech/
Multiformats
Self-describing data formats for interoperability
github.com/multiformats
IPLD
Linked data layer for distributed systems
github.com/ipld
Public AI
Open, verifiable AI infrastructure
publicai.co/
SAP BTP & Joule Team
Enterprise AI agent observability
Davos Economic Forum Committee: New Delhi
Policy & global security dialogue
Diagnosity
Health-AI observability pilots
field deployment
Park Pro Token on Ledger
  • Live testing on Park Pro Token (PPT) using a Ledger hardware wallet
  • PPT is listed across 10+ key EVM chains on SushiSwap and Mainnet
  • Real-world wallet, signing and cross-chain leakage telemetry feeds the detector
EVM · SushiSwap · Mainnet · Ledger
10: sunstone × leakdetect ai

Real-time leakage detection for AI agents.

Sunstone: a solar-powered IoT emergency pendant for elderly individuals living alone: is a key government-affiliated project that has added LeakDetect AI to its roadmap. The same temporal-anomaly principle Sunstone uses to catch a medical emergency from continuous sensor data is being applied to live AI-agent behavioral streams on Ethereum.

what gap does this address

LeakDetect AI today profiles a system, trains on its traces, and measures leakage: ideal for controlled hardware devices and TLS servers. AI agents executing autonomous transactions on Ethereum are different: they emit signals continuously, and an adversary won't wait for a profiling phase to finish. The Sunstone roadmap brings a streaming detector that learns normal behavior in place and raises a flag the moment something drifts.

Phase 1
Understanding & simulating agent behavioral signals

Identify the behavioral signals AI agents emit: transaction timing, gas usage per operation, signing behavior, API call sequences, response-time variance: and structure them as time-series streams. Synthesize a standardized Ethereum-agent dataset of normal sequences plus injected leakage patterns, in the same spirit as ASCAD / AES_HD for hardware. Statistically profile the normal sequences to establish a baseline the live detector can reason against.

Phase 2
Baseline temporal model

Validate the core idea: can a temporal model watching a live stream of agent behavioral signals detect leakage-shaped deviations? A Bidirectional LSTM consumes a sliding window of readings, trains only on normal sequences, and flags anything outside that envelope. BiLSTM reads context in both directions, which helps separate true anomalies from noise. Output: a Python PoC that ingests a simulated stream, learns normal, and raises real-time flags as new data arrives.

Phase 3
Attention over the temporal model

Different channels carry different signal depending on the leak: timing for timing leaks, gas for computation-heavy leaks. An attention layer on top of the BiLSTM learns per-channel and per-timestep weights, focusing the model on the most informative slices of the sequence. The weighted summary feeds the anomaly scoring head. Benchmarked against the Phase 2 baseline on detection accuracy and false-positive rate.

Phase 4
Integration with LeakDetect AI

The temporal detector sits alongside the existing profiling pipeline as a live monitoring layer: offline benchmarking stays, real-time flagging is added. A lightweight collection interface captures agent signals as time-series windows. Anomaly scores are mapped onto the existing VS / KRP / ILR vulnerability framework so outputs are consistent. Initial benchmarks on simulated Ethereum-agent scenarios give comparable numbers to the hardware and TLS suites already in the repo.

// collaborate with us

In decentralized systems, anything observable can eventually become exploitable.

We invite Ethereum researchers, wallet providers, AI-agent developers, infrastructure builders, and multi-chain teams to collaborate with LeakDetect AI in advancing the next generation of Ethereum security.