Feat(#12): URL 검사 엔드포인트 구현 및 하이브리드 검사 파이프라인 조립 - #16
Conversation
📝 WalkthroughWalkthroughChangesThe PR adds URL scanning contracts and utilities, refactors security integrations into engine classes, introduces hybrid URL analysis, and integrates URL maliciousness into the URL and smishing analysis
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant AnalyzeEndpoint
participant ScanService
participant URLTracker
participant VirusTotalEngine
participant GoogleSafeBrowsingEngine
AnalyzeEndpoint->>ScanService: scan_message_text(message)
ScanService->>URLTracker: extract_urls(message)
ScanService->>URLTracker: trace_url(original_url)
ScanService->>VirusTotalEngine: scan_url(traced_url)
ScanService->>GoogleSafeBrowsingEngine: scan_url(traced_url)
VirusTotalEngine-->>ScanService: normalized result
GoogleSafeBrowsingEngine-->>ScanService: normalized result
ScanService-->>AnalyzeEndpoint: URLScanResponse
Possibly related issues
Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 8
🧹 Nitpick comments (5)
app/dto/schemas.py (1)
19-31: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valuePrefer
model_config = ConfigDict(json_schema_extra=...)over the deprecatedclass Config.Pydantic v2 still honors the inner
Configclass but marks it deprecated; migrating avoids future breakage and silences the RUF012 false positive on this attribute.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@app/dto/schemas.py` around lines 19 - 31, The schema’s inner Config class is deprecated under Pydantic v2. Replace it with a model_config assignment using ConfigDict and preserve the existing json_schema_extra example unchanged, including its field values.Source: Linters/SAST tools
app/service/scan_service.py (2)
80-83: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winReplace
These
logger.debug(...)(or remove) before release.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@app/service/scan_service.py` around lines 80 - 83, Replace the four raw response print statements in the scan response output block with the module’s logger.debug calls, preserving the VirusTotal and Safe Browsing response details while avoiding unconditional stdout output. Remove the separator prints or include equivalent context in a single debug log.
87-97: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick win
"error" in str(vt_result)is a fragile way to detect engine failures.Stringifying the whole result and substring-checking for
errorcan both miss and misfire (e.g. a URL/status text containing "error"). Since the engines return exceptions (viareturn_exceptions=True) or a normalized dict, prefer explicit checks:isinstance(x, Exception)orx.get("error")/"error" in xon the dict directly.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@app/service/scan_service.py` around lines 87 - 97, Update the VirusTotal and Google Safe Browsing failure checks in the scan flow to avoid substring matching against str(vt_result) or str(gsb_result). Detect failures only when the result is an Exception or a dictionary containing an "error" key, while preserving the existing error logging, fallback results, and error_logs updates..vscode/settings.json (1)
2-5: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueWindows-only interpreter paths in committed editor config.
.venv/Scripts/python.exeand.venv/Lib/site-packagesare Windows layouts; on macOS/Linux these are.venv/bin/pythonand.venv/lib/pythonX.Y/site-packages. Committing machine-specific.vscode/settings.jsonwill misconfigure other contributors—consider gitignoring it.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.vscode/settings.json around lines 2 - 5, Remove the machine-specific Python interpreter and analysis paths from the committed .vscode/settings.json, or move them into a user-local ignored configuration so contributors on macOS, Linux, and Windows are not affected. Preserve only shared editor settings in version control and ensure the workspace no longer forces the Windows-specific .venv paths.app/core/config.py (1)
9-10: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueDrop the deprecated
env=kwargs
pydantic-settingsv2 ignoresField(..., env=...); these values are already picked up from the matching field names. Remove the extra kwargs, or switch tovalidation_aliasonly if the env var names need to differ.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@app/core/config.py` around lines 9 - 10, Update the VIRUSTOTAL_API_KEY and GOOGLE_SAFE_BROWSING_API_KEY field declarations to remove the deprecated env= arguments, relying on their matching field names for environment-variable loading; use validation_alias only if either environment name differs from its field name.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@app/main.py`:
- Around line 11-17: Update the CORSMiddleware configuration to use explicit
known gateway origin(s) instead of allow_origins=["*"] while retaining
allow_credentials=True, or remove allow_credentials when credentialed requests
are not required. Keep the existing methods and headers configuration unchanged.
In `@app/router/analyze.py`:
- Around line 48-49: Update the exception handler in the analyze route to log
the caught exception server-side instead of interpolating str(e) into the client
response. Return a generic Korean error message without internal details, and
configure ApiResponse.error or the route response handling to signal failure
with an appropriate non-200 status code while preserving the existing success
behavior.
In `@app/service/scan_service.py`:
- Around line 117-120: Replace the broad ".ru" substring check in the malicious
URL guard with a parsed-host suffix check using traced_url, so only hostnames
ending in ".ru" trigger the rule; preserve the existing testsafebrowsing
condition and malicious-state updates.
- Around line 69-78: Update the exception handler around the asyncio.wait_for
call in the hybrid scan flow to catch both asyncio.TimeoutError and the built-in
TimeoutError. Preserve the existing timeout log message and {"error": "Timeout"}
fallback assignments for vt_result and gsb_result.
In `@app/service/security/google_safe_browsing.py`:
- Line 11: Update the HTTP failure logging in the Google Safe Browsing request
flow to avoid stringifying HTTPStatusError, since its message contains the
API-key-bearing request URL. Log only safe details such as the response status
code and request method, or omit the exception text entirely; keep the existing
error handling behavior unchanged.
In `@app/service/security/mock_provider.py`:
- Around line 30-58: Align the mock security API used by parser.py with
mock_provider.py: expose the imported is_mock_enabled and get_mock_security_data
helpers from this module, or update parser.py to call the existing
MockSecurityEngine API. Ensure the parser import path resolves successfully and
preserves the current mock URL-scanning behavior.
In `@app/utils/url_tracker.py`:
- Around line 42-46: Bound the total execution time of trace_url, rather than
relying only on the per-request timeout in the HEAD/GET redirect loop. Add an
overall deadline or timeout around the tracing flow, including redirect handling
and GET fallback, so it cannot exceed the pipeline SLA before
scan_service.scan_message_text proceeds.
- Around line 29-46: Update trace_url to validate the initial URL and every
redirect target before issuing HEAD or GET requests, rejecting loopback,
private, link-local, and metadata addresses. Enforce the intended overall
timeout within trace_url itself so redirect processing cannot consume
substantially longer than the external asyncio.wait_for guard.
---
Nitpick comments:
In @.vscode/settings.json:
- Around line 2-5: Remove the machine-specific Python interpreter and analysis
paths from the committed .vscode/settings.json, or move them into a user-local
ignored configuration so contributors on macOS, Linux, and Windows are not
affected. Preserve only shared editor settings in version control and ensure the
workspace no longer forces the Windows-specific .venv paths.
In `@app/core/config.py`:
- Around line 9-10: Update the VIRUSTOTAL_API_KEY and
GOOGLE_SAFE_BROWSING_API_KEY field declarations to remove the deprecated env=
arguments, relying on their matching field names for environment-variable
loading; use validation_alias only if either environment name differs from its
field name.
In `@app/dto/schemas.py`:
- Around line 19-31: The schema’s inner Config class is deprecated under
Pydantic v2. Replace it with a model_config assignment using ConfigDict and
preserve the existing json_schema_extra example unchanged, including its field
values.
In `@app/service/scan_service.py`:
- Around line 80-83: Replace the four raw response print statements in the scan
response output block with the module’s logger.debug calls, preserving the
VirusTotal and Safe Browsing response details while avoiding unconditional
stdout output. Remove the separator prints or include equivalent context in a
single debug log.
- Around line 87-97: Update the VirusTotal and Google Safe Browsing failure
checks in the scan flow to avoid substring matching against str(vt_result) or
str(gsb_result). Detect failures only when the result is an Exception or a
dictionary containing an "error" key, while preserving the existing error
logging, fallback results, and error_logs updates.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: d6d4d5fe-1839-4635-b4aa-367203f83bb2
📒 Files selected for processing (15)
.vscode/settings.jsonapp/core/config.pyapp/dto/schemas.pyapp/main.pyapp/router/analyze.pyapp/router/analyze_voice.pyapp/service/scan_service.pyapp/service/security/base.pyapp/service/security/gemini_text_analyzer.pyapp/service/security/google_safe_browsing.pyapp/service/security/mock_provider.pyapp/service/security/virustotal.pyapp/service/url/extractor.pyapp/service/url/tracer.pyapp/utils/url_tracker.py
💤 Files with no reviewable changes (3)
- app/service/url/extractor.py
- app/service/url/tracer.py
- app/router/analyze_voice.py
| app.add_middleware( | ||
| CORSMiddleware, | ||
| allow_origins=["*"], | ||
| allow_credentials=True, | ||
| allow_methods=["*"], | ||
| allow_headers=["*"], | ||
| ) |
There was a problem hiding this comment.
🔒 Security & Privacy | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '\n== app/main.py ==\n'
cat -n app/main.py | sed -n '1,120p'
printf '\n== CORS-related references ==\n'
rg -n "CORSMiddleware|allow_origins|allow_credentials|allow_methods|allow_headers" -S .
printf '\n== Dependency files ==\n'
git ls-files | rg '(^|/)(requirements|pyproject|poetry\.lock|Pipfile|Pipfile\.lock|setup\.py|setup\.cfg|package\.json|uv\.lock)$' || trueRepository: SafeFam/SafeFam_AI
Length of output: 1221
🌐 Web query:
Starlette CORSMiddleware allow_origins "*" allow_credentials behavior docs
💡 Result:
As of July 2026, the configuration of Starlette's CORSMiddleware with both allow_origins=[""] and allow_credentials=True is treated as a security misconfiguration and is strictly prohibited [1]. In recent versions of Starlette, initializing CORSMiddleware with allow_origins=[""] and allow_credentials=True simultaneously will raise a ValueError at application startup [1]. This change aligns with W3C CORS specifications, which prohibit the use of credentials with a wildcard origin [1]. Prior to this enforcement, earlier versions of Starlette attempted to handle this combination by dynamically echoing the requesting Origin in the Access-Control-Allow-Origin response header instead of returning the literal * wildcard, whenever credentials (such as cookies or Authorization headers) were present [2][3]. However, this behavior was considered unreliable and potentially dangerous, leading to the current requirement for explicit origin specification when credentials are enabled [1][2]. To support credentialed requests, you must explicitly list the allowed origins (e.g., allow_origins=["https://example.com"]) rather than using the wildcard [1][4].
Citations:
- 1: fix(cors): raise ValueError when allow_credentials=True with allow_origins=["*"] Kludex/starlette#3246
- 2: Return explicit origin in CORS response when credentials are allowed Kludex/starlette#3137
- 3: Set explicit Origin in CORS preflight response if allow_credentials is True and allow_origins is wildcard Kludex/starlette#1113
- 4: https://fastapi.tiangolo.com/tutorial/cors/
Use explicit origins when credentials are enabled. allow_origins=["*"] with allow_credentials=True is invalid for credentialed CORS and will be rejected by current Starlette/FastAPI releases at startup. Restrict allow_origins to the known gateway origin(s), or drop allow_credentials if cookies/auth headers aren’t needed.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@app/main.py` around lines 11 - 17, Update the CORSMiddleware configuration to
use explicit known gateway origin(s) instead of allow_origins=["*"] while
retaining allow_credentials=True, or remove allow_credentials when credentialed
requests are not required. Keep the existing methods and headers configuration
unchanged.
| except Exception as e: | ||
| return ApiResponse.error(message=f"통합 스미싱 탐지 중 서버 에러가 발생했습니다: {str(e)}") |
There was a problem hiding this comment.
🔒 Security & Privacy | 🟡 Minor | ⚡ Quick win
Avoid returning raw exception text to clients.
f"...{str(e)}" can expose internal details (stack context, paths, keys embedded in messages) to callers, and the response still carries HTTP 200 from the route decorator. Log the exception server-side and return a generic message; consider signalling failure via status code as well.
🧰 Tools
🪛 Ruff (0.15.21)
[warning] 48-48: Do not catch blind exception: Exception
(BLE001)
[warning] 49-49: Use explicit conversion flag
Replace with conversion flag
(RUF010)
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@app/router/analyze.py` around lines 48 - 49, Update the exception handler in
the analyze route to log the caught exception server-side instead of
interpolating str(e) into the client response. Return a generic Korean error
message without internal details, and configure ApiResponse.error or the route
response handling to signal failure with an appropriate non-200 status code
while preserving the existing success behavior.
| try: | ||
| results = await asyncio.wait_for( | ||
| asyncio.gather(vt_task, gsb_task, return_exceptions=True), | ||
| timeout=6.0 | ||
| ) | ||
| vt_result, gsb_result = results | ||
| except asyncio.TimeoutException: | ||
| logger.error("[Pipeline Timeout] 외부 보안 API 응답 시간 초과로 하이브리드 스캔이 강제 타임아웃 처리되었습니다.") | ||
| vt_result = {"error": "Timeout"} | ||
| gsb_result = {"error": "Timeout"} |
There was a problem hiding this comment.
🩺 Stability & Availability | 🔴 Critical | ⚡ Quick win
🧩 Analysis chain
🌐 Web query:
python asyncio wait_for raises TimeoutError or TimeoutException
💡 Result:
When asyncio.wait_for reaches its specified timeout, it raises a TimeoutError exception [1][2]. It is important to note that the specific class of exception raised has changed over Python versions: 1. In Python 3.11 and later: asyncio.wait_for raises the built-in TimeoutError [3][2]. The name asyncio.TimeoutError has been retained as an alias for the built-in TimeoutError [3][4]. 2. In Python 3.10 and earlier: asyncio.wait_for raises asyncio.TimeoutError, which was a distinct exception class unrelated to the built-in TimeoutError [1][5][6]. If you are writing code that needs to support older versions of Python, you should catch asyncio.TimeoutError [7]. If you are working exclusively in Python 3.11+, you can catch either the built-in TimeoutError or the alias asyncio.TimeoutError [4]. There is no exception named TimeoutException in the Python standard library's asyncio module; TimeoutError is the correct exception to handle [1][3][2]. Additionally, if the task being awaited raises its own exception during the cancellation process triggered by a timeout, asyncio.wait_for will propagate that exception [1][8][2]. This behavior ensures that errors occurring during task cleanup are not silently suppressed [8][9].
Citations:
- 1: https://docs.python.org/3.10/library/asyncio-task.html
- 2: https://docs.python.org/3/library/asyncio-task.html
- 3: https://docs.python.org/3/library/asyncio-exceptions.html
- 4: https://stackoverflow.com/questions/72988580/python-asyncio-timouterror-is-not-using-builtin-timeouterror-as-base-class
- 5: https://docs.python.org/3.9/library/asyncio-task.html
- 6: https://docs.python.org/release/3.10.6/library/asyncio-api-index.html
- 7: https://stackoverflow.com/questions/74510354/exception-has-occurred-timeouterror-exception-no-description
- 8: python/cpython@382a563
- 9: asyncio.wait_for should reraise future exception even if timeout expires python/cpython#84787
🏁 Script executed:
#!/bin/bash
set -euo pipefail
sed -n '55,100p' app/service/scan_service.py | cat -nRepository: SafeFam/SafeFam_AI
Length of output: 2748
🏁 Script executed:
python3 - <<'PY'
import asyncio
print("has TimeoutException:", hasattr(asyncio, "TimeoutException"))
print("TimeoutError:", asyncio.TimeoutError)
try:
print("MRO:", asyncio.TimeoutError.mro())
except Exception as e:
print(type(e), e)
# Simulate except clause evaluation when attribute missing
class Dummy:
pass
try:
raise TimeoutError("boom")
except getattr(Dummy, "Missing", None):
print("caught")
except Exception as e:
print("handler exception:", type(e).__name__, e)
# Check whether built-in TimeoutError catches asyncio.TimeoutError
try:
raise asyncio.TimeoutError("async boom")
except TimeoutError as e:
print("builtin caught asyncio.TimeoutError:", type(e).__name__, e)
PYRepository: SafeFam/SafeFam_AI
Length of output: 653
🏁 Script executed:
python3 - <<'PY'
import asyncio
def probe():
try:
raise TimeoutError("boom")
except asyncio.TimeoutException:
return "caught"
except Exception as e:
return f"inner caught {type(e).__name__}: {e}"
try:
print(probe())
except Exception as e:
print(f"outer caught {type(e).__name__}: {e}")
PYRepository: SafeFam/SafeFam_AI
Length of output: 238
Catch the real timeout exception — asyncio.TimeoutException doesn’t exist, so a timeout raises AttributeError while matching this handler and skips the {"error": "Timeout"} fallback. Use except (asyncio.TimeoutError, TimeoutError): here.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@app/service/scan_service.py` around lines 69 - 78, Update the exception
handler around the asyncio.wait_for call in the hybrid scan flow to catch both
asyncio.TimeoutError and the built-in TimeoutError. Preserve the existing
timeout log message and {"error": "Timeout"} fallback assignments for vt_result
and gsb_result.
| if not is_malicious and (".ru" in traced_url or "testsafebrowsing" in traced_url): | ||
| logger.warning("[Infrastructure Guard] 외부 API 응답 공백 감지 - 로컬 정밀 위협 룰셋에 의해 악성 URL로 강제 전환합니다.") | ||
| is_malicious = True | ||
| risk_score = 0.75 |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | ⚡ Quick win
".ru" in traced_url is an over-broad substring match causing false positives.
Substring matching flags any URL whose text contains .ru, including legitimate hosts like docs.ruby-lang.org (.ru inside .ruby) or paths/query strings. This forces is_malicious=True incorrectly. Match on the parsed host suffix instead (e.g. urlparse(traced_url).hostname.endswith(".ru")).
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@app/service/scan_service.py` around lines 117 - 120, Replace the broad ".ru"
substring check in the malicious URL guard with a parsed-host suffix check using
traced_url, so only hostnames ending in ".ru" trigger the rule; preserve the
existing testsafebrowsing condition and malicious-state updates.
| class GoogleSafeBrowsingEngine(BaseSecurityEngine): | ||
| def __init__(self): | ||
| self.api_key = settings.GOOGLE_SAFE_BROWSING_API_KEY | ||
| self.api_url = f"https://safebrowsing.googleapis.com/v4/threatMatches:find?key={self.api_key}" |
There was a problem hiding this comment.
🔒 Security & Privacy | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '\n== file outline ==\n'
ast-grep outline app/service/security/google_safe_browsing.py --view expanded || true
printf '\n== relevant lines ==\n'
sed -n '1,140p' app/service/security/google_safe_browsing.pyRepository: SafeFam/SafeFam_AI
Length of output: 3158
🏁 Script executed:
#!/bin/bash
set -euo pipefail
python3 - <<'PY'
import inspect, httpx
print(httpx.__version__)
print(inspect.getsource(httpx.HTTPStatusError.__str__))
PYRepository: SafeFam/SafeFam_AI
Length of output: 1042
🏁 Script executed:
#!/bin/bash
set -euo pipefail
python3 - <<'PY'
import httpx
req = httpx.Request("POST", "https://example.com/path?key=SECRET123")
resp = httpx.Response(403, request=req)
try:
resp.raise_for_status()
except httpx.HTTPStatusError as e:
print(str(e))
print("URL:", e.request.url)
PYRepository: SafeFam/SafeFam_AI
Length of output: 367
Don't stringify HTTPStatusError here
httpx.HTTPStatusError.__str__() includes the request URL, and self.api_url embeds ?key=..., so logger.error(... {str(e)}) will leak GOOGLE_SAFE_BROWSING_API_KEY into logs on HTTP failures. Log only the status code/method, or omit the exception string.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@app/service/security/google_safe_browsing.py` at line 11, Update the HTTP
failure logging in the Google Safe Browsing request flow to avoid stringifying
HTTPStatusError, since its message contains the API-key-bearing request URL. Log
only safe details such as the response status code and request method, or omit
the exception text entirely; keep the existing error handling behavior
unchanged.
| async def scan_url(self, url: str) -> dict: | ||
| """ | ||
| [상용 인터페이스 구현] | ||
| 가상 DB를 매칭하여 상용 엔진과 동일한 규격의 딕셔너리를 반환합니다. | ||
| """ | ||
| logger.info(f"🚨 [MOCK SECURITY ENGINE] Sandbox API 우회 매칭 -> {url}") | ||
|
|
||
| # URL이 DB에 없으면 안전한 상태인 default_safe 적용 | ||
| data = self.mock_db.get(url, self.default_safe) | ||
|
|
||
| gsb_malicious = data.get("google_safe_browsing", False) | ||
| vt_malicious_count = data.get("malicious_count", 0) | ||
|
|
||
| # 악성 여부 및 위험도 점수 가중치 산정 (상용 규격과 통일) | ||
| is_malicious = gsb_malicious or (vt_malicious_count >= 3) | ||
|
|
||
| if gsb_malicious: | ||
| raw_score = 0.95 | ||
| elif vt_malicious_count > 0: | ||
| raw_score = min(0.2 + (vt_malicious_count * 0.08), 1.0) | ||
| else: | ||
| raw_score = 0.0 | ||
|
|
||
| def is_mock_enabled() -> bool: | ||
| return MOCK_ENABLED | ||
| return { | ||
| "is_malicious": is_malicious, | ||
| "raw_score": round(raw_score, 2), | ||
| "detected_count": vt_malicious_count, | ||
| "status": "completed" | ||
| } |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
rg -nP '\bMockSecurityEngine\b' --type=pyRepository: SafeFam/SafeFam_AI
Length of output: 156
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '\n== mock_provider.py ==\n'
if [ -f app/service/security/mock_provider.py ]; then
sed -n '1,220p' app/service/security/mock_provider.py
else
echo 'missing app/service/security/mock_provider.py'
fi
printf '\n== scan_service.py ==\n'
if [ -f app/service/security/scan_service.py ]; then
sed -n '1,260p' app/service/security/scan_service.py
else
echo 'missing app/service/security/scan_service.py'
fi
printf '\n== references ==\n'
rg -n "MockSecurityEngine|MOCK_ENABLED|vt_malicious_count|google_safe_browsing" app/service/security -g '*.py'Repository: SafeFam/SafeFam_AI
Length of output: 5026
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '\n== security tree ==\n'
find app/service/security -maxdepth 2 -type f | sort
printf '\n== parser.py ==\n'
sed -n '1,220p' app/service/security/parser.py
printf '\n== mock provider usages ==\n'
rg -n "mock_provider|MockSecurityEngine|get_mock_text_analysis_data|MOCK_SECURITY_API|MOCK_ENABLED" . -g '*.py'Repository: SafeFam/SafeFam_AI
Length of output: 4552
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '\n== mock_provider.py (numbered) ==\n'
cat -n app/service/security/mock_provider.py
printf '\n== scan_service.py (numbered) ==\n'
cat -n app/service/scan_service.py
printf '\n== symbol search ==\n'
rg -n "def (is_mock_enabled|get_mock_security_data)|class MockSecurityEngine|from app\.service\.security\.mock_provider import" app/service -g '*.py'Repository: SafeFam/SafeFam_AI
Length of output: 11281
🏁 Script executed:
#!/bin/bash
set -euo pipefail
rg -n "from app\.service\.security\.parser import|import app\.service\.security\.parser|security\.parser|analyze_threat_pipeline|determine_final_threat_grade" . -g '*.py'Repository: SafeFam/SafeFam_AI
Length of output: 2042
mock_provider.py is missing the helpers parser.py imports.
app/service/security/parser.py imports is_mock_enabled and get_mock_security_data, but this module only defines MockSecurityEngine and get_mock_text_analysis_data. That breaks the parser import path; either export the expected helpers here or update parser.py to use the current API.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@app/service/security/mock_provider.py` around lines 30 - 58, Align the mock
security API used by parser.py with mock_provider.py: expose the imported
is_mock_enabled and get_mock_security_data helpers from this module, or update
parser.py to call the existing MockSecurityEngine API. Ensure the parser import
path resolves successfully and preserves the current mock URL-scanning behavior.
| async def trace_url(url: str, max_redirects: int = 5, timeout: float = 3.0) -> str: | ||
|
|
||
| current_url = url | ||
|
|
||
| limits = httpx.Limits(max_keepalive_connections=5, max_connections=10) | ||
|
|
||
| async with httpx.AsyncClient(limits=limits, follow_redirects=False) as client: | ||
| for attempt in range(max_redirects): | ||
| try: | ||
| headers = { | ||
| "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36" | ||
| } | ||
|
|
||
| response = await client.head(current_url, headers=headers, timeout=timeout) | ||
|
|
||
| # HEAD를 차단하거나 거부하는 서버(400, 404, 405)에 대응하기 위한 GET 폴백 | ||
| if response.status_code in [400, 404, 405]: | ||
| response = await client.get(current_url, headers=headers, timeout=timeout) |
There was a problem hiding this comment.
🔒 Security & Privacy | 🟠 Major | 🏗️ Heavy lift
Block internal targets before following redirects. trace_url sends HEAD/GET requests to user-controlled URLs and follows Location headers, so a crafted link or redirect can reach loopback, RFC1918, link-local, or metadata endpoints. It also runs before the later asyncio.wait_for, so this path can still consume ~30s before the 6s guard applies.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@app/utils/url_tracker.py` around lines 29 - 46, Update trace_url to validate
the initial URL and every redirect target before issuing HEAD or GET requests,
rejecting loopback, private, link-local, and metadata addresses. Enforce the
intended overall timeout within trace_url itself so redirect processing cannot
consume substantially longer than the external asyncio.wait_for guard.
| response = await client.head(current_url, headers=headers, timeout=timeout) | ||
|
|
||
| # HEAD를 차단하거나 거부하는 서버(400, 404, 405)에 대응하기 위한 GET 폴백 | ||
| if response.status_code in [400, 404, 405]: | ||
| response = await client.get(current_url, headers=headers, timeout=timeout) |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win
Total tracing time is unbounded relative to the pipeline SLA.
Worst case is max_redirects × (HEAD + GET) × timeout (~30s), and trace_url runs before the 6s asyncio.wait_for guard in scan_service.scan_message_text, so the endpoint's overall latency isn't bounded. Consider an overall deadline for tracing or wrapping it in a timeout as well.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@app/utils/url_tracker.py` around lines 42 - 46, Bound the total execution
time of trace_url, rather than relying only on the per-request timeout in the
HEAD/GET redirect loop. Add an overall deadline or timeout around the tracing
flow, including redirect handling and GET fallback, so it cannot exceed the
pipeline SLA before scan_service.scan_message_text proceeds.
1 participant
📝 개요
Spring Boot 메인 서버와 연동하기 위한 독립된 URL 검사 API 엔드포인트를 구축했습니다.
문자 본문 내 URL 추출부터 단축 URL의 실제 주소 추적, 그리고
VirusTotal과Google Safe BrowsingAPI의 비동기 병렬 호출 및 동기화까지의 전체 분석 파이프라인을 조립하여 실시간 프로던셕(Real-API) 모드를 완성했습니다.🔗 관련 이슈
🎯 주요 변경 사항
Spring Boot 통신용 URL 검사 요청/응답 DTO 및 엔드포인트 구축
URL 추출 및 단축 URL 실제 주소 추적 모듈 구현
bit.ly등 단축 URL 뒤에 숨겨진 실제 악성 행선지 주소를 끝까지 추적하는 비동기 리다이렉트 헬퍼 함수(trace_url) 매핑하이브리드 URL 검사 비동기 파이프라인 및 예외 처리 가드 구축
VirusTotal과Google Safe Browsing인프라를 6.0초 타임아웃 제한 내에서 비동기 호출하도록 처리📸 사진
✅ PR 체크리스트
uvicorn구동 또는 테스트 코드)를 통과했습니다.Summary by CodeRabbit
New Features
Bug Fixes