mirror of
https://github.com/yogeshojha/rengine.git
synced 2026-10-21 07:41:28 +02:00
The export engine, with no surface on it yet. Given a dimension, a scope and the filter body a result table POSTs to search, it writes the same rows to csv, json or plain text. It selects only the columns the file will carry, so a vulnerability export does not drag a stored request and response through memory to throw them away, and it streams with yield_per, so a 50,000 row export holds one chunk at a time. Statement timeout is the export's own: a background job does not read under the api's twenty second ceiling. Two columns a table shows are joined per chunk rather than per row — the ports behind a host's addresses, and the hosts behind an address — because #523 and #703 ask for those by name. The four rollup counts the api computes per page are left out rather than written blank. A cell opening with = + - @ tab or return is a formula to Excel and Sheets, and page titles and banners are written by the target, so every cell is prefixed and stripped of control characters. CSV carries a BOM or Excel reads utf-8 as latin-1. Proven against a live scan and in tests: for every dimension, filtered or not, the export writes exactly as many rows as the table reports. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LYZfaNjq3Ym8GPvWYuHisS
122 lines
4.1 KiB
Python
122 lines
4.1 KiB
Python
"""Vulnerability statement building, shared by the api's search and the worker's export."""
|
|
|
|
from __future__ import annotations
|
|
|
|
from typing import TYPE_CHECKING
|
|
|
|
from sqlalchemy import case, cast, func, not_, or_, select
|
|
from sqlalchemy.dialects.postgresql import JSONB
|
|
|
|
from shared.definitions.asset_query import VULN_QUERY
|
|
from shared.definitions.vulnerabilities import ACTIONABLE_SEVERITIES, SEVERITY_ORDER
|
|
from shared.models.vulnerability import Vulnerability
|
|
from shared.services.asset_query import (
|
|
QueryScope,
|
|
VulnQueryContext,
|
|
compile_vuln_query,
|
|
parse_query,
|
|
vuln_corroborated,
|
|
vuln_corroborated_ids,
|
|
vuln_is_new,
|
|
vuln_state,
|
|
vuln_suppressed,
|
|
)
|
|
|
|
if TYPE_CHECKING:
|
|
from datetime import datetime
|
|
|
|
from shared.models.vulnerability import VulnerabilityFilter
|
|
|
|
|
|
def severity_rank():
|
|
return case(
|
|
{name: index for index, name in enumerate(SEVERITY_ORDER)},
|
|
value=Vulnerability.severity,
|
|
else_=len(SEVERITY_ORDER),
|
|
)
|
|
|
|
|
|
def context(scope: QueryScope, now: datetime) -> VulnQueryContext:
|
|
return VulnQueryContext(scope=scope, now=now)
|
|
|
|
|
|
def apply_filter(query, f: VulnerabilityFilter, scope: QueryScope):
|
|
if f.severities:
|
|
query = query.where(Vulnerability.severity.in_(f.severities))
|
|
elif not f.include_info:
|
|
query = query.where(Vulnerability.severity.in_(ACTIONABLE_SEVERITIES))
|
|
if f.protocols:
|
|
query = query.where(Vulnerability.protocol.in_(f.protocols))
|
|
if f.templates:
|
|
query = query.where(Vulnerability.template_id.in_(f.templates))
|
|
if f.hosts:
|
|
query = query.where(Vulnerability.host.in_(f.hosts))
|
|
if f.scanners:
|
|
query = query.where(Vulnerability.scanner.in_(f.scanners))
|
|
if f.tags:
|
|
query = query.where(
|
|
or_(
|
|
*[
|
|
func.jsonb_exists(cast(Vulnerability.tags, JSONB), tag)
|
|
for tag in f.tags
|
|
]
|
|
)
|
|
)
|
|
if f.kev:
|
|
query = query.where(Vulnerability.is_kev.is_(True))
|
|
if f.cve:
|
|
query = query.where(
|
|
func.jsonb_array_length(cast(Vulnerability.cve_ids, JSONB)) > 0
|
|
)
|
|
if f.new:
|
|
query = query.where(vuln_is_new(scope))
|
|
if f.corroborated:
|
|
query = query.where(vuln_corroborated(scope))
|
|
if f.states:
|
|
query = query.where(vuln_state(scope).in_(f.states))
|
|
elif not f.include_suppressed:
|
|
query = query.where(not_(vuln_suppressed(scope)))
|
|
return query
|
|
|
|
|
|
def order(query, f: VulnerabilityFilter, scope: QueryScope):
|
|
if f.sort == "risk":
|
|
confirmed = vuln_corroborated_ids(scope).subquery()
|
|
return query.outerjoin(confirmed, confirmed.c.id == Vulnerability.id).order_by(
|
|
severity_rank().asc(),
|
|
Vulnerability.is_kev.desc(),
|
|
Vulnerability.exploit_score.desc(),
|
|
confirmed.c.id.isnot(None).desc(),
|
|
Vulnerability.epss_score.desc().nulls_last(),
|
|
Vulnerability.cvss_score.desc().nulls_last(),
|
|
Vulnerability.template_id.asc(),
|
|
Vulnerability.matched_at.asc(),
|
|
)
|
|
column = {
|
|
"severity": severity_rank(),
|
|
"name": Vulnerability.template_name,
|
|
"template": Vulnerability.template_id,
|
|
"host": Vulnerability.host,
|
|
"seen": Vulnerability.discovered_at,
|
|
"cvss": Vulnerability.cvss_score,
|
|
"epss": Vulnerability.epss_score,
|
|
"exploit": Vulnerability.exploit_score,
|
|
"type": Vulnerability.protocol,
|
|
}.get(f.sort, severity_rank())
|
|
primary = column.desc() if f.order == "desc" else column.asc()
|
|
return query.order_by(
|
|
primary.nulls_last(),
|
|
Vulnerability.template_id.asc(),
|
|
Vulnerability.matched_at.asc(),
|
|
)
|
|
|
|
|
|
def scoped(scope: QueryScope, f: VulnerabilityFilter, columns=None):
|
|
base = select(Vulnerability) if columns is None else select(*columns)
|
|
base = base.where(scope.match(Vulnerability.scan_id))
|
|
return apply_filter(base, f, scope)
|
|
|
|
|
|
def compiled(scope: QueryScope, f: VulnerabilityFilter, now: datetime):
|
|
return compile_vuln_query(parse_query(f.q, VULN_QUERY), context(scope, now))
|