MIESC Roadmap

July 14, 2026 · View on GitHub

Development roadmap for MIESC. The authoritative history of shipped changes lives in CHANGELOG.md; this file looks forward.

Current Version: 6.0.0 (released 2026-07-13)

MIESC is feature-complete and in active maintenance. Highlights of the 5.x line:

  • Detection: 9 defense layers, 50 integrated tools; SmartBugs-curated recall 95.8% (Paper 1 reproducible profile; v-next 99.3% pending baseline, see Near-term); EVMBench 92.5% with a multi-provider ensemble.
  • Remediation (miesc fix): self-contained find→fix→verify pipeline (Paper 2).
  • RAG false-positive filtering, cross-tool Bayesian confidence, agentic deep-audit.
  • Formal bridge (miesc verify): Certora/Scribble/SMTChecker spec generation.
  • Multi-chain: EVM + Cardano/Plutus-Aiken, Algorand/TEAL, Stellar/Soroban, Starknet/Cairo.
  • Research: miesc evaluate framework, plugin system, two academic papers.

Versions 5.2.0 → 6.0.0 are shipped (see CHANGELOG). An earlier version of this roadmap listed them as future work; corrected 2026-06.

Near-term — Papers v-next + hardening

  • Paper 1 v-next baseline: detection recall 95.8%→99.3% (deterministic detectors close 5 of 6 prior misses). Evidence ready; baseline pending final PDF build and freeze re-tag.
  • Paper 2 v-next: raise the external-validation clean rate (v2 baseline: 70/123 standalone-Slither clean of HIGH, 88/123 eliminate on re-scan) and lean further on independent verification (Slither/SMTChecker/Foundry on patched contracts) to remove re-scan circularity. Standalone compilation is already 123/123 in v2.
  • Reproducibility: the benchmark harness reports external-tool crash counts per run; document the unloaded-machine measurement protocol.
  • Quality: 88% line coverage, 75% mutation score on core v6 modules, and a deterministic order-independent suite; ongoing robustness hardening toward higher coverage.

Medium-term

  • Promote additional chain analyzers (Solana, Move/CosmWasm) from experimental to beta.
  • Expand the RAG knowledge base and per-detector calibration.
  • Deeper LLM-ensemble consensus and confidence calibration.

Longer-term

The v6.0 platform-enablement work — CI-native finding baseline & suppression, SARIF inline PR annotations, editor-agnostic code-action + LSP diagnostics contracts, unified formal-verification report (Scribble/Kontrol), versioned plugin API + conformance suite + reference plugins, and a webhook/Slack notifier — shipped in 6.0.0 (see Completed). Beyond the medium-term items above:

  • Deepen formal verification (more provers; counterexample→PoC linkage).
  • Grow the community plugin ecosystem on the versioned plugin API + conformance suite.
  • Editor and SaaS product surfaces consume these stable core contracts; those UIs are built separately from this open-source core.

Completed milestones (summary)

Full detail in CHANGELOG.md. Historical version-specific plans are archived under docs/archive/.

  • 6.0.0 — CI-native adoption (finding baseline & suppression, fail-only-on-new), SARIF inline PR annotations, editor-agnostic code-action output + miesc lsp diagnostics server, unified formal-verification report (Scribble/Kontrol), versioned plugin API + conformance suite + reference plugins, webhook/Slack notifier. 88% coverage, 75% mutation, DPGA 9/9 (self-assessed, application under review).
  • 5.4.x — remediation pipeline (Paper 2), research evaluation framework, multi-chain analyzers (Cardano/Algorand/Stellar), PyPI + multi-arch Docker.
  • 5.1–5.3.x — RAG (59→ patterns) with caching, CLI refactor, plugin system, multi-chain alpha→beta.
  • 4.x (2025) — 9 defense layers, MCP support, ML-based false-positive reduction.

How to influence the roadmap

Open a GitHub Discussion or a feature request. See GOVERNANCE.md for the decision process.