Deserialization of Untrusted Data (CWE-502)

Deserialization of Untrusted Data

The product deserializes untrusted data without sufficiently verifying that the resulting data will be valid.

Many programming languages allow the serialization of objects for storage or transmission. When untrusted data is deserialized, it can lead to code execution, denial of service, or other unintended consequences.

普及度

中

3 言語をカバー

影響度

クリティカル

3 件の重大度クリティカルなルール

予防

文書化済み

7 件の修正例

2 予防

この脆弱性の修正方法

7 件の Shoulder 検出ルールに基づく Deserialization of Untrusted Data の予防策。

🐹 Go すべて表示 Go 詳細 →

Insecure Deserialization HIGH

Use strict typed structs instead of interface{} and avoid gob with untrusted data

+16 -10 go

  package main
  
  import (
-     "encoding/gob"
-     "net/http"
- )
- 
- func handler(w http.ResponseWriter, r *http.Request) {
-     // Vulnerable: gob decoding untrusted HTTP body
-     dec := gob.NewDecoder(r.Body)
-     var data interface{}
-     if err := dec.Decode(&data); err != nil {
-         http.Error(w, err.Error(), 400)
+     "encoding/json"
+     "net/http"
+ )
+ 
+ type UserRequest struct {
+     Name  string `json:"name"`
+     Email string `json:"email"`
+ }
+ 
+ func handler(w http.ResponseWriter, r *http.Request) {
+     // Safe: typed struct with JSON (data-only, no code execution)
+     var req UserRequest
+     dec := json.NewDecoder(r.Body)
+     dec.DisallowUnknownFields()
+     if err := dec.Decode(&req); err != nil {
+         http.Error(w, "Invalid request", 400)
          return
      }
  }

LLM Training Data Poisoning HIGH

Validate all training data against strict schemas and apply content moderation before ingestion

+12 -0 go

  func indexHandler(w http.ResponseWriter, r *http.Request) {
      var docs []Document
      json.NewDecoder(r.Body).Decode(&docs)
+ 
+     validate := validator.New()
+     for _, doc := range docs {
+         if err := validate.Struct(doc); err != nil {
+             http.Error(w, "validation failed", http.StatusBadRequest)
+             return
+         }
+         if flagged, _ := moderationCheck(doc.Content); flagged {
+             http.Error(w, "content policy violation", http.StatusBadRequest)
+             return
+         }
+     }
      vectorDB.Upsert(docs)
  }

🟨 JavaScript すべて表示 JavaScript 詳細 →

LLM Training Data Poisoning HIGH

Validate training data against schemas and use content moderation before fine-tuning

+4 -2 javascript

  app.post('/finetune', async (req, res) => {
-   await openai.files.create({
-     file: req.body.trainingData,
+   const validated = trainingSchema.parse(req.body.trainingData);
+   const moderated = await moderateContent(validated);
+   await openai.files.create({
+     file: moderated,
      purpose: 'fine-tune'
    });
  });

Unsafe Deserialization CRITICAL

Use JSON.parse() instead of node-serialize, and yaml.SAFE_SCHEMA for YAML parsing

+10 -8 javascript

  const express = require('express');
- const serialize = require('node-serialize');
- const app = express();
- 
- app.post('/restore', (req, res) => {
-   const sessionData = req.body.session;
-   const session = serialize.deserialize(sessionData);
-   req.session = session;
-   res.json({ restored: true });
+ const app = express();
+ 
+ app.post('/restore', (req, res) => {
+   try {
+     const session = JSON.parse(req.body.session);
+     req.session = session;
+     res.json({ restored: true });
+   } catch (e) {
+     res.status(400).json({ error: 'Invalid session data' });
+   }
  });

🐍 Python すべて表示 Python 詳細 →

LLM Training Data Poisoning HIGH

Validate training data with Pydantic schemas and apply content moderation before ingestion

+21 -4 python

- @app.route('/finetune', methods=['POST'])
- def finetune():
-     training_data = request.json['data']
-     client.files.create(file=training_data, purpose='fine-tune')
+ from pydantic import BaseModel, validator
+ 
+ class TrainingExample(BaseModel):
+     prompt: str
+     completion: str
+ 
+     @validator('prompt', 'completion')
+     def validate_length(cls, v):
+         if len(v) > 4000:
+             raise ValueError('Content too long')
+         return v
+ 
+ @app.route('/finetune', methods=['POST'])
+ async def finetune():
+     examples = [TrainingExample(**ex) for ex in request.json['data']]
+     moderation = await openai.moderations.create(
+         input=[ex.completion for ex in examples]
+     )
+     if any(r.flagged for r in moderation.results):
+         return {'error': 'Content policy violation'}, 400
+     client.files.create(file=json.dumps([ex.dict() for ex in examples]), purpose='fine-tune')
      return {'status': 'queued'}

Unsafe Deserialization CRITICAL

Replace pickle/marshal with JSON or other safe serialization formats

+7 -7 python

- import pickle
- from flask import request
- 
- @app.route('/load', methods=['POST'])
- def load():
-     data = request.get_data()
-     obj = pickle.loads(data)
+ import json
+ from flask import request
+ 
+ @app.route('/load', methods=['POST'])
+ def load():
+     data = request.get_data()
+     obj = json.loads(data)
      return str(obj)

Unsafe YAML Deserialization CRITICAL

Use yaml.safe_load() instead of yaml.load() to prevent code execution

+1 -1 python

  import yaml
  
  def parse_config(yaml_string):
-     config = yaml.load(yaml_string)
+     config = yaml.safe_load(yaml_string)
      return config

3 検出

コードの脆弱性を見つける

Shoulderを使用してコードのDeserialization of Untrusted Dataパターンをスキャンしましょう。 7 ルール.

ターミナル

# Scan with Shoulder CLI
npx @shoulderdev/cli trust --cwe=502

# Or scan entire project
npx @shoulderdev/cli trust .

検出ルール (7)

🐍 Python 3 rules

LLM Training Data Poisoning HIGH

Detects untrusted or unvalidated data flowing into AI/LLM fine-tuning or training processes. OWASP LLM03 - Training Data Poisoning. Training data poisoning can: - Introduce backdoors into model behavior - Bias model outputs maliciously - Embed harmful content that appears in responses - Compromise model accuracy and reliability - Create security vulnerabilities in model behavior

Unsafe Deserialization CRITICAL

Detects untrusted user input being deserialized using unsafe methods like pickle.loads() or yaml.load().

Unsafe YAML Deserialization CRITICAL

Detects unsafe YAML deserialization using yaml.load() without SafeLoader.

🐹 Go 2 rules

Insecure Deserialization HIGH

Detects truly dangerous deserialization in Go. Unlike Java or Python, Go's encoding/json is safe (data-only parsing, no code execution). This rule focuses on: - gob.Decoder: Can instantiate arbitrary types, potential RCE (CRITICAL) - json/yaml/xml to interface{}: Type confusion risk when combined with untrusted input (MEDIUM) Note: json.Unmarshal to typed structs is NOT flagged as it cannot execute code.

LLM Training Data Poisoning HIGH

Detects untrusted data flowing into AI/LLM fine-tuning or training processes without validation.

🟨 Javascript 2 rules

LLM Training Data Poisoning HIGH

Unsafe Deserialization CRITICAL

Detects user input flowing to unsafe deserialization functions like node-serialize or yaml.load().

🔷 Typescript 2 rules

LLM Training Data Poisoning HIGH

Unsafe Deserialization CRITICAL

Detects user input flowing to unsafe deserialization functions like node-serialize or yaml.load().

4 警告サイン

コードレビューで注目すべき点

これらのパターンはDeserialization of Untrusted Dataの潜在的な脆弱性を示しています。コードレビューとセキュリティ監査中に探してください。

🟠

Untrusted data is deserialized without validation go-insecure-deserialization

🟠

truly dangerous deserialization in Go go-insecure-deserialization

🟠

Untrusted data flows to ... without validation go-llm-training-data-poisoning

🟠

untrusted data flowing into AI/LLM fine-tuning or training processes without validation go-llm-training-data-poisoning

🟠

untrusted or unvalidated data flowing into AI/LLM fine-tuning or training processes javascript-llm-training-data-poisoning

🔴

user input flowing to unsafe deserialization functions like node-serialize or yaml javascript-unsafe-deserialization

🔴

untrusted user input being deserialized using unsafe methods like pickle python-unsafe-deserialization

🔴

unsafe YAML deserialization using yaml python-yaml-deserialization

🔍

コードベースをスキャン: Deserialization of Untrusted Data

Shoulder CLI はコードベース全体から脆弱なパターンを見つけます。

npx shoulder trust すべてのルールを表示