feat: implement Pillars 2-4 and fix a production false-positive in Pillar 1

Pillar 1 v1.1 (shipped earlier) flagged benign browsing bursts as
beaconing: on short captures, grid periods longer than the observed
span never complete a cycle, so significance saturates to 1 and the
score sticks at exactly 0.5. v1.2 caps the grid at a third of the
observed span, requires a minimum number of completed cycles, and
excludes protocols that are periodic by nature (NTP, DNS, mDNS,
ICMP). Validated against 300 seeds per scenario: false positives on
bursty traffic drop from the 290-300/300 that v1.1 produced to 0/300,
with every beacon-detection scenario scoring unchanged.

Add Pillar 2 (persistent/long-lived sessions, with Zeek's fragmented
TCP sessions rejoined after a 5-minute gap), Pillar 3 (DNS tunneling
via Shannon-entropy information volume per subdomain per hour; the
Jensen-Shannon divergence was measured and does not separate tunnels
from benign CDN subdomains, so it stays in the evidence at zero
weight instead of in the score), and Pillar 4 (deterministic
noisy-OR fusion across the three pillars).

The fusion formula was validated against 8 labeled real-traffic
vectors (IoT-23, CTU-Normal) with tolerance <= 1e-4. A known
limitation is kept, not hidden: a benign Philips HUE bulb's strict
periodic polling and a 21.9h session score it above three real C2
hosts. No weight or threshold was tuned to mask this; it is
documented in the architecture docs instead.

Also fixes real bugs found along the way: ingest_dir's partition
count summed file names instead of distinct dates, DuckDB errors
were counted as bad rows, /api/alerts always returned an empty list
regardless of the last analysis, and the dashboard still claimed
every pillar was unimplemented.

Restores four design corrections that were missing from
docs/architecture.md (KL to Jensen-Shannon, the persistent-session
detector, the JA4 licensing boundary, and deterministic reduction
order), and catalogs the public datasets used for validation
(IoT-23, CTU-Normal, Mendeley DTQBC) with their licenses and SHA-256
checksums under research/validation/.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VAyr8d7HuAdUGB6aA5xmDD
This commit is contained in:
Lucas RayanandClaude Sonnet 5 committed 2026-11-01 00:28:54 -03:00
1 parent 8a80504c45
commit c64bbbf775
41 files changed
+4034 -348

No files matched your search

+15 -15
View File
@@ -41,7 +41,7 @@
<div class="side-foot">
<div class="status-row">
<span id="connection" class="pill"><span class="dot"></span><span id="connectionText">Conectando…</span></span>
<span class="pill muted">esqueleto</span>
<span class="pill muted">pré-lançamento</span>
</div>
<div class="status-row">
<span id="version" class="mono" data-i18n-skip>—</span>
@@ -60,9 +60,9 @@
<button id="settingsBtn" type="button" class="icon-btn" aria-label="Configurações" title="Configurações">⚙</button>
</header>
<section class="notice notice-warn">
<strong>Aviso de escopo.</strong>
<span>Este é um esqueleto pré-implementação. Os quatro pilares matemáticos, o parser Zeek e o armazenamento DuckDB ainda não estão implementados.</span>
<section class="notice">
<strong>Estado atual.</strong>
<span>Ingestão de conn.log e dns.log do Zeek, armazenamento Parquet/DuckDB e os pilares 1 a 4 em versão inicial, validados com tráfego público real. Ainda não: ingestão contínua automática, alertas persistidos entre reinícios, movimentação lateral e sessões UDP.</span>
</section>
<section class="notice">
@@ -74,26 +74,26 @@
<article class="card">
<span class="card-kicker">Pilar 1</span>
<h2>Beaconing</h2>
<p>Periodograma de Rayleigh/Schuster. Em desenvolvimento.</p>
<span class="tag">em desenvolvimento</span>
<p>Periodograma de Rayleigh/Schuster com grade limitada ao intervalo observado.</p>
<span class="tag">versão inicial</span>
</article>
<article class="card">
<span class="card-kicker">Pilar 2</span>
<h2>Sessões persistentes</h2>
<p>Detector de conexões longas. Em desenvolvimento.</p>
<span class="tag">em desenvolvimento</span>
<p>Sessões TCP longas, inclusive as que o Zeek divide por inatividade.</p>
<span class="tag">versão inicial</span>
</article>
<article class="card">
<span class="card-kicker">Pilar 3</span>
<h2>Anomalias de protocolo</h2>
<p>Features determinísticas por protocolo. Em desenvolvimento.</p>
<span class="tag">em desenvolvimento</span>
<p>Volume de informação (entropia de Shannon) em subdomínios DNS.</p>
<span class="tag">versão inicial</span>
</article>
<article class="card">
<span class="card-kicker">Pilar 4</span>
<h2>Fusão de evidências</h2>
<p>Jensen-Shannon e prova por alerta. Em desenvolvimento.</p>
<span class="tag">em desenvolvimento</span>
<p>Prioridade por host com evidência por alerta.</p>
<span class="tag">versão inicial</span>
</article>
</section>
@@ -119,16 +119,16 @@
<div class="panel-head">
<div>
<h2>Alertas recentes</h2>
<p class="hint">Cada alerta futuro carregará a evidência determinística que o sustenta.</p>
<p class="hint">Cada alerta carrega a evidência determinística que o sustenta.</p>
</div>
</div>
<ul id="alertList" class="alert-list">
<li class="empty">Nenhum alerta ainda. O motor de detecção será ligado nas próximas specs.</li>
<li class="empty">Nenhum alerta ainda. Analise um diretório de logs do Zeek.</li>
</ul>
</section>
<footer class="foot">
<span>Acauã · esqueleto pré-implementação</span>
<span>Acauã · pré-lançamento</span>
<span class="mono">MIT OR Apache-2.0</span>
</footer>
</main>