AarthShah/CyberQwen-14B-Security-V2
Text Generation • Updated • 1
id stringlengths 15 17 | category stringclasses 59
values | difficulty stringclasses 8
values | quality_score float64 5 98 | evidence_status stringclasses 4
values | instruction stringlengths 50 303 | input stringlengths 45 10.1k | output stringlengths 439 4.01k | vulnerability_class stringclasses 58
values | synthetic bool 1
class | reasoning_skill stringclasses 11
values | verification_method stringclasses 11
values | metadata dict |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
CYBER_M_FINAL_001 | Authentication and authorization | Advanced | 98 | Verified | Triage and harden this Node.js (jsonwebtoken) implementation against production attacks:
```javascript
const jwt = require('jsonwebtoken'); | function createToken(userId) {
return jwt.sign({ sub: userId }, 'my_secret_key_12345', { expiresIn: '1h' });
}
```
Provide vulnerability details, attack vector, secure fix, and Semgrep rule. | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_002 | Authentication and authorization | Advanced | 98 | Verified | Security verification required: Does this Node.js (jsonwebtoken) snippet have any OWASP Top 10 vulnerabilities?
```javascript
const jwt = require('jsonwebtoken'); | function createToken(userId) {
return jwt.sign({ sub: userId }, 'my_secret_key_12345', { expiresIn: '1h' });
}
```
Provide vulnerability details, attack vector, secure fix, and Semgrep rule. | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_003 | Authentication and authorization | Advanced | 98 | Verified | Our team used Cursor / GitHub Copilot to write this Node.js (jsonwebtoken) code. Review it for vulnerabilities:
```javascript
const jwt = require('jsonwebtoken'); | function createToken(userId) {
return jwt.sign({ sub: userId }, 'my_secret_key_12345', { expiresIn: '1h' });
}
```
Provide vulnerability details, attack vector, secure fix, and Semgrep rule. | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_004 | Authentication and authorization | Advanced | 93 | Verified | [AI CODE AUDIT REQUEST: Node.js (jsonwebtoken)]
Review this AI-generated code written for our web application:
```javascript
const jwt = require('jsonwebtoken'); | function createToken(userId) {
return jwt.sign({ sub: userId }, 'my_secret_key_12345', { expiresIn: '1h' });
}
```
Perform a comprehensive enterprise security audit:
1) Identify the security flaw & CWE
2) Explain how an attacker exploits this
3) Provide the secure, production-hardened replacement code
4) Output a Sem... | ### 1. Vulnerability Assessment
* **Category**: Node.js (jsonwebtoken)
* **Vulnerability**: Hardcoded JWT Secret Key
* **CWE**: CWE-798
### 2. Exploit Mechanics
Hardcoding the secret key `'my_secret_key_12345'` in source code allows anyone with repository access or public git leaks to sign valid JWTs and forge admin t... | null | null | null | null | null |
CYBER_M_FINAL_005 | Authentication and authorization | Advanced | 98 | Verified | Audit this AI-written Node.js (jsonwebtoken) component for security flaws before deployment:
```javascript
const jwt = require('jsonwebtoken'); | function createToken(userId) {
return jwt.sign({ sub: userId }, 'my_secret_key_12345', { expiresIn: '1h' });
}
```
Provide vulnerability details, attack vector, secure fix, and Semgrep rule. | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_006 | Root-cause vulnerability analysis | Intermediate | 98 | Verified | [AI CODE AUDIT REQUEST: React Frontend]
Review this AI-generated code written for our web application:
```javascript
import React from 'react'; | export function BlogPost({ postContent }) {
return (
<article className="prose">
<div dangerouslySetInnerHTML={{ __html: postContent }} />
</article>
);
}
```
Perform a comprehensive enterprise security audit:
1) Identify the security flaw & CWE
2) Explain how an attacker exploits this
3) Provide the ... | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_007 | Root-cause vulnerability analysis | Intermediate | 98 | Verified | Triage and harden this React Frontend implementation against production attacks:
```javascript
import React from 'react'; | export function BlogPost({ postContent }) {
return (
<article className="prose">
<div dangerouslySetInnerHTML={{ __html: postContent }} />
</article>
);
}
```
Provide vulnerability details, attack vector, secure fix, and Semgrep rule. | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_008 | Root-cause vulnerability analysis | Intermediate | 98 | Verified | Audit this AI-written React Frontend component for security flaws before deployment:
```javascript
import React from 'react'; | export function BlogPost({ postContent }) {
return (
<article className="prose">
<div dangerouslySetInnerHTML={{ __html: postContent }} />
</article>
);
}
```
Provide vulnerability details, attack vector, secure fix, and Semgrep rule. | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_009 | Root-cause vulnerability analysis | Intermediate | 98 | Verified | Security verification required: Does this React Frontend snippet have any OWASP Top 10 vulnerabilities?
```javascript
import React from 'react'; | export function BlogPost({ postContent }) {
return (
<article className="prose">
<div dangerouslySetInnerHTML={{ __html: postContent }} />
</article>
);
}
```
Provide vulnerability details, attack vector, secure fix, and Semgrep rule. | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_010 | Root-cause vulnerability analysis | Intermediate | 98 | Verified | Our team used Cursor / GitHub Copilot to write this React Frontend code. Review it for vulnerabilities:
```javascript
import React from 'react'; | export function BlogPost({ postContent }) {
return (
<article className="prose">
<div dangerouslySetInnerHTML={{ __html: postContent }} />
</article>
);
}
```
Provide vulnerability details, attack vector, secure fix, and Semgrep rule. | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_011 | Source-to-sink and data-flow analysis | Intermediate | 93 | Verified | Trace the complete data flow of this vulnerability from the untrusted input source through the system to the security-sensitive sink. Include a data flow diagram. | Vulnerability ID: GO-2026-6209
Source: osv
Title: Traefik: Gateway API route identity collision allows cross-namespace backend hijacking in github.com/traefik/traefik
Description: Traefik: Gateway API route identity collision allows cross-namespace backend hijacking in github.com/traefik/traefik
Package: github.com/tra... | DATA FLOW TRACE: GO-2026-6209
============================================================
This traces untrusted data from entry to exploitation.
PHASE 1 — SOURCE (where untrusted data enters):
Sources: Any user-controllable input including HTTP parameters,
headers, file contents, environment variables, IPC messa... | null | null | null | null | null |
CYBER_M_FINAL_012 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Perform a root cause analysis of this vulnerability. What is the fundamental programming error, and why did it occur? | Vulnerability ID: GO-2026-6209
Source: osv
Title: Traefik: Gateway API route identity collision allows cross-namespace backend hijacking in github.com/traefik/traefik
Description: Traefik: Gateway API route identity collision allows cross-namespace backend hijacking in github.com/traefik/traefik
Package: github.com/tra... | ### Technical Security Analysis
ROOT CAUSE ANALYSIS: GO-2026-6209
============================================================
WHAT WENT WRONG:
Traefik: Gateway API route identity collision allows cross-namespace backend hijacking in github.com/traefik/traefik
ROOT CAUSE IDENTIFICATION:
The root cause is related to ... | null | null | null | null | null |
CYBER_M_FINAL_013 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Perform rigorous evidence validation, verifying every conclusion against provided code, patch, and advisory metadata. | Vulnerability ID: GO-2026-6209
Source: osv
Title: Traefik: Gateway API route identity collision allows cross-namespace backend hijacking in github.com/traefik/traefik
Description: Traefik: Gateway API route identity collision allows cross-namespace backend hijacking in github.com/traefik/traefik
Package: github.com/tra... | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_014 | Root-cause vulnerability analysis | Intermediate | 94 | Verified | Audit this Dockerfile for production container security anti-patterns. | FROM ubuntu:22.04
RUN apt-get update && apt-get install -y python3 curl sudo
COPY app.py /app/app.py
EXPOSE 8080
CMD ["python3", "/app/app.py"] | ### Container Image Architecture Review
1. **Root-Cause Analysis:**
The container lacks a `USER` directive (runs as root), includes unnecessary sensitive utilities (`sudo`, `curl`), and does not clean up apt package caches, expanding attack surface in case of remote execution.
2. **Remediation:**
Use a minimal ... | null | null | null | null | null |
CYBER_M_FINAL_015 | Root-cause vulnerability analysis | Beginner | 97 | Verified | Review format string vulnerability in debug logging handler. | // --- sys_debug.c ---
void log_event(const char *user_input) {
printf(user_input);
} | ### Uncontrolled Format String Flaw
1. **Root-Cause Analysis:**
Passing user-controlled strings as the format argument to `printf` allows `%x`, `%s`, and `%n` specifiers to read arbitrary stack/heap memory or overwrite memory addresses.
2. **Remediation:**
Always use explicit format specifiers: `printf("%s", us... | null | null | null | null | null |
CYBER_M_FINAL_016 | Patch correctness and regression analysis | Intermediate | 96 | Verified | Evaluate this security patch for Path Traversal to determine if it can be bypassed. | // Patch: file = file.replace(/\.\.\//g, ''); | ### Patch Correctness Audit: Incomplete Sanitization
1. **Verdict:**
FAILED / BYPASSABLE FIX.
2. **Bypass Mechanics:**
Single-pass replacement is bypassed by nested sequences: `....//` collapses to `../`. Furthermore, fails to block Windows backslash separators (`..\`).
3. **Remediation:**
Use canonical pat... | null | null | null | null | null |
CYBER_M_FINAL_017 | Source-to-sink and data-flow analysis | Intermediate | 93 | Verified | Trace the complete data flow of this vulnerability from the untrusted input source through the system to the security-sensitive sink. Include a data flow diagram. | Vulnerability ID: GHSA-838g-gr43-qqg9
Source: osv
Title: PyLoad vulnerable to Path Traversal via Package Folder Name in set_package_data
Description: ### Summary
No sanitization of package folder name allows writing files anywhere outside the intended download directory.
#### Affected Component
- `src/pyload/core/api/... | DATA FLOW TRACE: GHSA-838g-gr43-qqg9
============================================================
This traces untrusted data from entry to exploitation.
PHASE 1 — SOURCE (where untrusted data enters):
Sources: File path parameters in URLs/APIs, uploaded filenames,
configuration values from external input.
PHASE ... | null | null | null | null | null |
CYBER_M_FINAL_018 | Business logic vulnerabilities | Intermediate | 98 | Verified | Evaluate multiple alternative root cause hypotheses and justify why the chosen explanation is best supported by evidence. | Vulnerability ID: GHSA-838g-gr43-qqg9
Source: osv
Title: PyLoad vulnerable to Path Traversal via Package Folder Name in set_package_data
Description: ### Summary
No sanitization of package folder name allows writing files anywhere outside the intended download directory.
#### Affected Component
- `src/pyload/core/api/... | ### Technical Security Analysis
ALTERNATIVE HYPOTHESIS EVALUATION: GHSA-838g-gr43-qqg9
============================================================
TARGET WEAKNESS: Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal') (CWE-22)
1. HYPOTHESIS A (Missing Precondition Check):
Evaluation: Bou... | null | null | null | null | null |
CYBER_M_FINAL_019 | SSRF, path traversal, injection, and other vulnerability classes | Intermediate | 98 | Verified | Generate a formal security vulnerability report for this advisory. | Vulnerability ID: GHSA-838g-gr43-qqg9
Source: osv
Title: PyLoad vulnerable to Path Traversal via Package Folder Name in set_package_data
Description: ### Summary
No sanitization of package folder name allows writing files anywhere outside the intended download directory.
#### Affected Component
- `src/pyload/core/api/... | ### Technical Security Analysis
============================================================
SECURITY VULNERABILITY REPORT: GHSA-838g-gr43-qqg9
============================================================
Severity: CVSS_V3
CWE: CWE-22 (Improper Limitation of a Pathname to a Restricted Directory ('Path Trav... | null | null | null | null | null |
CYBER_M_FINAL_020 | Secure architecture and design | Intermediate | 93 | Verified | Propose a secure architectural redesign that eliminates this vulnerability CLASS by design, not just this specific instance. Include code examples. | Vulnerability ID: GHSA-838g-gr43-qqg9
Source: osv
Title: PyLoad vulnerable to Path Traversal via Package Folder Name in set_package_data
Description: ### Summary
No sanitization of package folder name allows writing files anywhere outside the intended download directory.
#### Affected Component
- `src/pyload/core/api/... | SECURE DESIGN PROPOSAL: GHSA-838g-gr43-qqg9
============================================================
Instead of patching individual instances, redesign the affected
component to eliminate Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal') as a class of vulnerability.
DESIGN PATTERN: Vi... | null | null | null | null | null |
CYBER_M_FINAL_021 | SSRF, path traversal, injection, and other vulnerability classes | Intermediate | 98 | Verified | Write a Semgrep static analysis rule that would detect this vulnerability pattern in source code. The rule should be deployable in a CI/CD pipeline. | Vulnerability ID: GHSA-838g-gr43-qqg9
Source: osv
Title: PyLoad vulnerable to Path Traversal via Package Folder Name in set_package_data
Description: ### Summary
No sanitization of package folder name allows writing files anywhere outside the intended download directory.
#### Affected Component
- `src/pyload/core/api/... | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_022 | Advisory and vulnerability triage | Intermediate | 98 | Verified | Perform a full architecture security review centered on this vulnerability. Map the components, identify architectural weaknesses, and provide a repository-level review checklist. | Vulnerability ID: PYSEC-2021-42
Source: osv
Description: Pillow before 8.1.1 allows attackers to cause a denial of service (memory consumption) because the reported size of a contained image is not properly checked for an ICO container, and thus an attempted memory allocation can be very large.
Severity: MEDIUM
CWE: CW... | ### Technical Security Analysis
ARCHITECTURE SECURITY REVIEW: PYSEC-2021-42
============================================================
COMPONENT MAP:
[pillow] in [PyPI] (Python)
|
v
[External Input] --> [Processing] --> [Security-Sensitive Operation]
^ ^ ... | null | null | null | null | null |
CYBER_M_FINAL_023 | Memory safety | Intermediate | 98 | Verified | Evaluate multiple alternative root cause hypotheses and justify why the chosen explanation is best supported by evidence. | Vulnerability ID: PYSEC-2021-42
Source: osv
Description: Pillow before 8.1.1 allows attackers to cause a denial of service (memory consumption) because the reported size of a contained image is not properly checked for an ICO container, and thus an attempted memory allocation can be very large.
Severity: MEDIUM
CWE: CW... | ### Technical Security Analysis
ALTERNATIVE HYPOTHESIS EVALUATION: PYSEC-2021-42
============================================================
TARGET WEAKNESS: Uncontrolled Resource Consumption (CWE-400)
1. HYPOTHESIS A (Missing Precondition Check):
Evaluation: Boundary check absent at entry point to sensitive ope... | null | null | null | null | null |
CYBER_M_FINAL_024 | Memory safety | Intermediate | 98 | Verified | Generate a formal security vulnerability report for this advisory. | Vulnerability ID: PYSEC-2021-42
Source: osv
Description: Pillow before 8.1.1 allows attackers to cause a denial of service (memory consumption) because the reported size of a contained image is not properly checked for an ICO container, and thus an attempted memory allocation can be very large.
Severity: MEDIUM
CWE: CW... | ### Technical Security Analysis
============================================================
SECURITY VULNERABILITY REPORT: PYSEC-2021-42
============================================================
Severity: MEDIUM
CWE: CWE-400 (Uncontrolled Resource Consumption)
Package: pillow
Ecosystem: PyPI
Publ... | null | null | null | null | null |
CYBER_M_FINAL_025 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Perform a root cause analysis of this vulnerability. What is the fundamental programming error, and why did it occur? | Vulnerability ID: MAL-2023-6838
Source: osv
Title: Malicious code in superad (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: checkmarx (79b0668ffd8a4981cf6aedf730c404b3e7d92c8654629a4048aaefc37c48490c)
EsqueleSquad group published nearly 6000 malicious PyPi and NPM packages,... | ### Technical Security Analysis
ROOT CAUSE ANALYSIS: MAL-2023-6838
============================================================
WHAT WENT WRONG:
---
_-= Per source details. Do not edit below this line.=-_
## Source: checkmarx (79b0668ffd8a4981cf6aedf730c404b3e7d92c8654629a4048aaefc37c48490c)
EsqueleSquad group publ... | null | null | null | null | null |
CYBER_M_FINAL_026 | Uncertainty calibration and evidence discipline | Intermediate | 98 | Verified | Critically self-review generated security findings, identify weak assumptions, and correct unsupported claims. | Vulnerability ID: MAL-2023-6838
Source: osv
Title: Malicious code in superad (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: checkmarx (79b0668ffd8a4981cf6aedf730c404b3e7d92c8654629a4048aaefc37c48490c)
EsqueleSquad group published nearly 6000 malicious PyPi and NPM packages,... | ### Technical Security Analysis
SELF-REVIEW & ASSUMPTION AUDIT: MAL-2023-6838
============================================================
1. CRITICAL AUDIT OF INITIAL FINDINGS:
- Scope Check: Verified finding applies specifically to 'superad' within specified version bounds.
- Evidence Check: Grounded in publi... | null | null | null | null | null |
CYBER_M_FINAL_027 | Uncertainty calibration and evidence discipline | Intermediate | 98 | Verified | Perform frontier security research reasoning using the exact 12-step format: Understand -> Analyze -> Evidence -> Alternative Hypotheses -> Root Cause -> Security Principle -> Impact -> Fix -> Verification -> Self-Review -> Confidence -> Limitations. Ground all statements strictly in provided evidence. | Vulnerability ID: MAL-2023-6838
Source: osv
Title: Malicious code in superad (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: checkmarx (79b0668ffd8a4981cf6aedf730c404b3e7d92c8654629a4048aaefc37c48490c)
EsqueleSquad group published nearly 6000 malicious PyPi and NPM packages,... | STRUCTURED SECURITY REASONING: MAL-2023-6838
============================================================
-> UNDERSTAND:
Target package 'superad' (Python) is affected by a security flaw (Unclassified).
Advisory summary: Malicious code in superad (PyPI)
-> ANALYZE:
Analyzing advisory metadata and taxonomic weakness pr... | null | null | null | null | null |
CYBER_M_FINAL_028 | Uncertainty calibration and evidence discipline | Intermediate | 91 | Verified | Perform explicit uncertainty estimation, stating limitations and identifying missing information required for total certainty. | Vulnerability ID: MAL-2023-6838
Source: osv
Title: Malicious code in superad (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: checkmarx (79b0668ffd8a4981cf6aedf730c404b3e7d92c8654629a4048aaefc37c48490c)
EsqueleSquad group published nearly 6000 malicious PyPi and NPM packages,... | UNCERTAINTY ESTIMATION & BOUNDING: MAL-2023-6838
============================================================
1. KNOWN EVIDENCE (HIGH CERTAINTY):
- Vulnerability ID: MAL-2023-6838
- Weakness Class: None - None
2. UNKNOWN / INSIGHT GAPS (MEDIUM/LOW CERTAINTY):
- Complete multi-file call graph (source patch no... | null | null | null | null | null |
CYBER_M_FINAL_029 | Root-cause vulnerability analysis | Intermediate | 93 | Verified | Analyze the provided technical artifact, identify security vulnerabilities or alerts, and provide remediation. | Is this AI implementation secure against OWASP Top 10 exploits?
```
void log_msg(const char *user, const char *action) {
char buf[128];
sprintf(buf, "User %s performed %s\n", user, action);
write(fd, buf, strlen(buf));
}
```
Provide CWE, exploit vector, secure fix, and Semgrep rule. | ### Security Review: Stack Buffer Overflow via sprintf (CWE-120 / CWE-121)
**Exploit Risk**: Attacker supplies user or action > 128 bytes, corrupting the stack frame and hijacking return pointer.
**Remediated Implementation**:
```
void log_msg(const char *user, const char *action) {
char buf[128];
if (!user |... | null | null | null | null | null |
CYBER_M_FINAL_030 | Patch correctness and regression analysis | Intermediate | 93 | Verified | Analyze the provided technical artifact, identify security vulnerabilities or alerts, and provide remediation. | [AI CODE AUDIT REQUEST]
Review this AI-generated code snippet (C / C++):
```
void log_msg(const char *user, const char *action) {
char buf[128];
sprintf(buf, "User %s performed %s\n", user, action);
write(fd, buf, strlen(buf));
}
```
Perform a professional security audit: 1) Identify CWE & flaw, 2) Explain ... | ### 1. Security Analysis
* **Vulnerability**: Stack Buffer Overflow via sprintf
* **CWE**: CWE-120 / CWE-121
### 2. Exploit Mechanics
Attacker supplies user or action > 128 bytes, corrupting the stack frame and hijacking return pointer.
### 3. Hardened Production Code
```
void log_msg(const char *user, const char *ac... | null | null | null | null | null |
CYBER_M_FINAL_031 | Root-cause vulnerability analysis | Intermediate | 93 | Verified | Analyze the provided technical artifact, identify security vulnerabilities or alerts, and provide remediation. | Inspect this AI-generated code block before we push to production:
```
void log_msg(const char *user, const char *action) {
char buf[128];
sprintf(buf, "User %s performed %s\n", user, action);
write(fd, buf, strlen(buf));
}
```
Provide CWE, exploit vector, secure fix, and Semgrep rule. | ### Security Review: Stack Buffer Overflow via sprintf (CWE-120 / CWE-121)
**Exploit Risk**: Attacker supplies user or action > 128 bytes, corrupting the stack frame and hijacking return pointer.
**Remediated Implementation**:
```
void log_msg(const char *user, const char *action) {
char buf[128];
if (!user |... | null | null | null | null | null |
CYBER_M_FINAL_032 | Root-cause vulnerability analysis | Intermediate | 93 | Verified | Analyze the provided technical artifact, identify security vulnerabilities or alerts, and provide remediation. | Our developer used Copilot to write this function. Audit it for vulnerabilities:
```
void log_msg(const char *user, const char *action) {
char buf[128];
sprintf(buf, "User %s performed %s\n", user, action);
write(fd, buf, strlen(buf));
}
```
Provide CWE, exploit vector, secure fix, and Semgrep rule. | ### Security Review: Stack Buffer Overflow via sprintf (CWE-120 / CWE-121)
**Exploit Risk**: Attacker supplies user or action > 128 bytes, corrupting the stack frame and hijacking return pointer.
**Remediated Implementation**:
```
void log_msg(const char *user, const char *action) {
char buf[128];
if (!user |... | null | null | null | null | null |
CYBER_M_FINAL_033 | Root-cause vulnerability analysis | Intermediate | 93 | Verified | Analyze the provided technical artifact, identify security vulnerabilities or alerts, and provide remediation. | Perform static security triage on this pull request code snippet:
```
void log_msg(const char *user, const char *action) {
char buf[128];
sprintf(buf, "User %s performed %s\n", user, action);
write(fd, buf, strlen(buf));
}
```
Provide CWE, exploit vector, secure fix, and Semgrep rule. | ### Security Review: Stack Buffer Overflow via sprintf (CWE-120 / CWE-121)
**Exploit Risk**: Attacker supplies user or action > 128 bytes, corrupting the stack frame and hijacking return pointer.
**Remediated Implementation**:
```
void log_msg(const char *user, const char *action) {
char buf[128];
if (!user |... | null | null | null | null | null |
CYBER_M_FINAL_034 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Show the patch diff for this vulnerability. What exact code change was made, and what does the before/after comparison look like? | Vulnerability ID: GHSA-x4cm-m36h-c6qj
Source: osv
Title: Improper Input Validation in ansible
Description: An input validation vulnerability was found in Ansible's mysql_user module before 2.2.1.0, which may fail to correctly change a password in certain circumstances. Thus the previous password would still be active w... | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_035 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Generate a formal security vulnerability report for this advisory. | Vulnerability ID: GHSA-x4cm-m36h-c6qj
Source: osv
Title: Improper Input Validation in ansible
Description: An input validation vulnerability was found in Ansible's mysql_user module before 2.2.1.0, which may fail to correctly change a password in certain circumstances. Thus the previous password would still be active w... | ### Technical Security Analysis
============================================================
SECURITY VULNERABILITY REPORT: GHSA-x4cm-m36h-c6qj
============================================================
Severity: CVSS_V3
Package: ansible
Ecosystem: PyPI
Published: 2018-10-10T17:23:33Z
DESCRIPTION:
An in... | null | null | null | null | null |
CYBER_M_FINAL_036 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Mine and detect recurring security anti-patterns across multiple files, modules, and services in the repository. | Vulnerability ID: MAL-2024-11562
Source: osv
Title: Malicious code in cometlogger (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: kam193 (627a435be86fedd2eccd2f0af61fa8b19c8c12dd781fc1bdc064d836e1eff19a)
Multi-purpose infostealer, including grabbing cookies, files, informati... | ### Technical Security Analysis
REPOSITORY-WIDE PATTERN MINING: MAL-2024-11562 (cometlogger)
============================================================
1. RECURRING DEFENSIVE PATTERN AUDIT:
Auditing codebase for other instances of None.
2. SYSTEMIC VULNERABILITY SURFACE:
Vulnerabilities in utility or parsing... | null | null | null | null | null |
CYBER_M_FINAL_037 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Generate a formal security vulnerability report for this advisory. | Vulnerability ID: MAL-2024-11562
Source: osv
Title: Malicious code in cometlogger (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: kam193 (627a435be86fedd2eccd2f0af61fa8b19c8c12dd781fc1bdc064d836e1eff19a)
Multi-purpose infostealer, including grabbing cookies, files, informati... | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_038 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Compare and correlate findings across static analysis, dependency scanning, patch history, and configuration reviews. | Vulnerability ID: MAL-2024-11562
Source: osv
Title: Malicious code in cometlogger (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: kam193 (627a435be86fedd2eccd2f0af61fa8b19c8c12dd781fc1bdc064d836e1eff19a)
Multi-purpose infostealer, including grabbing cookies, files, informati... | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_039 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Write an executive summary of this vulnerability for non-technical leadership. Focus on business risk, impact, and recommended action. | Vulnerability ID: MAL-2024-11562
Source: osv
Title: Malicious code in cometlogger (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: kam193 (627a435be86fedd2eccd2f0af61fa8b19c8c12dd781fc1bdc064d836e1eff19a)
Multi-purpose infostealer, including grabbing cookies, files, informati... | ### Technical Security Analysis
EXECUTIVE SUMMARY: MAL-2024-11562
RISK: HIGH
ISSUE: A security vulnerability was identified in cometlogger that could allow an attacker to compromise affected systems.
BUSINESS IMPACT:
- Potential data breach, regulatory compliance violations
- Reputational damage, customer trust ero... | null | null | null | null | null |
CYBER_M_FINAL_040 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Perform rigorous evidence validation, verifying every conclusion against provided code, patch, and advisory metadata. | Vulnerability ID: MAL-2024-11562
Source: osv
Title: Malicious code in cometlogger (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: kam193 (627a435be86fedd2eccd2f0af61fa8b19c8c12dd781fc1bdc064d836e1eff19a)
Multi-purpose infostealer, including grabbing cookies, files, informati... | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_041 | Uncertainty calibration and evidence discipline | Intermediate | 98 | Verified | Critically self-review generated security findings, identify weak assumptions, and correct unsupported claims. | Vulnerability ID: MAL-2024-11562
Source: osv
Title: Malicious code in cometlogger (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: kam193 (627a435be86fedd2eccd2f0af61fa8b19c8c12dd781fc1bdc064d836e1eff19a)
Multi-purpose infostealer, including grabbing cookies, files, informati... | ### Technical Security Analysis
SELF-REVIEW & ASSUMPTION AUDIT: MAL-2024-11562
============================================================
1. CRITICAL AUDIT OF INITIAL FINDINGS:
- Scope Check: Verified finding applies specifically to 'cometlogger' within specified version bounds.
- Evidence Check: Grounded in ... | null | null | null | null | null |
CYBER_M_FINAL_042 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Explain this vulnerability to a developer in plain language. What happened, what should they do right now, and how can they prevent it in the future? | Vulnerability ID: MAL-2024-11562
Source: osv
Title: Malicious code in cometlogger (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: kam193 (627a435be86fedd2eccd2f0af61fa8b19c8c12dd781fc1bdc064d836e1eff19a)
Multi-purpose infostealer, including grabbing cookies, files, informati... | ### Technical Security Analysis
DEVELOPER WALKTHROUGH: MAL-2024-11562
============================================================
Plain-language explanation for developers who need to understand
and fix this vulnerability.
WHAT HAPPENED (in plain English):
---
_-= Per source details. Do not edit below this line.=-... | null | null | null | null | null |
CYBER_M_FINAL_043 | Uncertainty calibration and evidence discipline | Intermediate | 98 | Verified | Perform frontier security research reasoning using the exact 12-step format: Understand -> Analyze -> Evidence -> Alternative Hypotheses -> Root Cause -> Security Principle -> Impact -> Fix -> Verification -> Self-Review -> Confidence -> Limitations. Ground all statements strictly in provided evidence. | Vulnerability ID: MAL-2024-11562
Source: osv
Title: Malicious code in cometlogger (PyPI)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: kam193 (627a435be86fedd2eccd2f0af61fa8b19c8c12dd781fc1bdc064d836e1eff19a)
Multi-purpose infostealer, including grabbing cookies, files, informati... | STRUCTURED SECURITY REASONING: MAL-2024-11562
============================================================
-> UNDERSTAND:
Target package 'cometlogger' (Python) is affected by a security flaw (Unclassified).
Advisory summary: Malicious code in cometlogger (PyPI)
-> ANALYZE:
Analyzing advisory metadata and taxonomic we... | null | null | null | null | null |
CYBER_M_FINAL_044 | Uncertainty calibration and evidence discipline | Intermediate | 91 | Verified | Perform explicit uncertainty estimation, stating limitations and identifying missing information required for total certainty. | Vulnerability ID: MAL-2025-108529
Source: osv
Title: Malicious code in senior_dove_z3n (npm)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: amazon-inspector (926adcb7fe66c447876595af6899fbef71e45ae4674b3e02e4ccfdfaecd4a567)
This package appears to be part of the tea.xyz token rewa... | UNCERTAINTY ESTIMATION & BOUNDING: MAL-2025-108529
============================================================
1. KNOWN EVIDENCE (HIGH CERTAINTY):
- Vulnerability ID: MAL-2025-108529
- Weakness Class: None - None
2. UNKNOWN / INSIGHT GAPS (MEDIUM/LOW CERTAINTY):
- Complete multi-file call graph (source patc... | null | null | null | null | null |
CYBER_M_FINAL_045 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Mine and detect recurring security anti-patterns across multiple files, modules, and services in the repository. | Vulnerability ID: MAL-2025-108529
Source: osv
Title: Malicious code in senior_dove_z3n (npm)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: amazon-inspector (926adcb7fe66c447876595af6899fbef71e45ae4674b3e02e4ccfdfaecd4a567)
This package appears to be part of the tea.xyz token rewa... | ### Technical Security Analysis
REPOSITORY-WIDE PATTERN MINING: MAL-2025-108529 (senior_dove_z3n)
============================================================
1. RECURRING DEFENSIVE PATTERN AUDIT:
Auditing codebase for other instances of None.
2. SYSTEMIC VULNERABILITY SURFACE:
Vulnerabilities in utility or pa... | null | null | null | null | null |
CYBER_M_FINAL_046 | Patch correctness and regression analysis | Intermediate | 98 | Verified | Compare and correlate findings across static analysis, dependency scanning, patch history, and configuration reviews. | Vulnerability ID: MAL-2025-108529
Source: osv
Title: Malicious code in senior_dove_z3n (npm)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: amazon-inspector (926adcb7fe66c447876595af6899fbef71e45ae4674b3e02e4ccfdfaecd4a567)
This package appears to be part of the tea.xyz token rewa... | ### Technical Vulnerability Analysis & Remediation
1. **Root-Cause Analysis:**
Vulnerability arises from insufficient input validation and unsafe parameter handling at the execution sink.
2. **Attack Vector & Exploitability Preconditions:**
- **Attack Path:** An attacker crafts malformed input specifically targ... | null | null | null | null | null |
CYBER_M_FINAL_047 | Patch correctness and regression analysis | Intermediate | 88 | Verified | Analyze performance vs security trade-offs and justify secure engineering design choices. | Vulnerability ID: MAL-2025-108529
Source: osv
Title: Malicious code in senior_dove_z3n (npm)
Description:
---
_-= Per source details. Do not edit below this line.=-_
## Source: amazon-inspector (926adcb7fe66c447876595af6899fbef71e45ae4674b3e02e4ccfdfaecd4a567)
This package appears to be part of the tea.xyz token rewa... | PERFORMANCE VS SECURITY TRADE-OFF ANALYSIS: MAL-2025-108529
============================================================
TARGET DEFECT: None
1. NAIVE PERFORMANCE OPTIMIZATION (INSECURE):
Omitting bounds checks or schema validations in hot execution loops to save CPU cycles.
2. SECURITY IMPACT OF BYPASS:
Bypass... | null | null | null | null | null |
CYBER_M_FINAL_048 | False-positive identification | Intermediate | 98 | Verified | Perform a comprehensive security code review of this Python function. Determine whether any security vulnerabilities are present, identify the security controls implemented, and provide a final assessment. | Component ID: BENIGN-CMD-01
Language: Python
Ecosystem: PyPI
Domain: Command Injection Prevention
```
import subprocess
def inspect_archive(archive_path: str):
"""Lists archive contents without invoking system shell."""
# Safe: arguments passed as list with shell=False
cmd = ['tar', '-tzf', archive_path]
... | ### Security Code Review: Negative Control (Benign Implementation)
SECURITY CODE REVIEW: BENIGN-CMD-01
============================================================
1. AUDIT VERDICT: SECURE (NO VULNERABILITY DETECTED)
- Severity: NONE
- Weakness Analyzed: Command Injection Prevention
2. TECHNICAL JUSTIFICATION:... | null | null | null | null | null |
CYBER_M_FINAL_049 | Source-to-sink and data-flow analysis | Intermediate | 98 | Verified | Audit this pull request change in Python for potential security regressions or unsafe practices before merging. | Pull Request Review: PR-BENIGN-CMD-01
Language: Python
Feature Area: Command Injection Prevention
```
import subprocess
def inspect_archive(archive_path: str):
"""Lists archive contents without invoking system shell."""
# Safe: arguments passed as list with shell=False
cmd = ['tar', '-tzf', archive_path]
... | ### Technical Security Analysis
PULL REQUEST SECURITY AUDIT: PR-BENIGN-CMD-01
============================================================
1. AUDIT STATUS: APPROVED (CLEAN)
- Vulnerabilities Found: 0
- Regression Risk: NEGLIGIBLE
2. CODE QUALITY & DEFENSIVE CONTROLS:
- Invokes the binary directly via execve... | null | null | null | null | null |
CYBER_M_FINAL_050 | Race conditions and concurrency | Intermediate | 97 | Verified | Examine this Python file update logic for Time-of-Check to Time-of-Use (TOCTOU) file race conditions. | # --- config_updater.py ---
def update_shared_config(new_data):
if os.path.exists('/etc/app/config.json'):
with open('/etc/app/config.json', 'w') as f:
f.write(new_data) | ### Non-Atomic File Truncation & TOCTOU Race
1. **Root-Cause Analysis:**
Opening an existing shared file directly with `'w'` mode truncates the file to 0 bytes before writing new data. Concurrent reader processes accessing `/etc/app/config.json` during the write window observe an empty or truncated JSON file, throw... | null | null | null | null | null |
CYBER_M_FINAL_051 | Memory safety | Advanced | 98 | Verified | Perform frontier security research reasoning using the exact 12-step format: Understand -> Analyze -> Evidence -> Alternative Hypotheses -> Root Cause -> Security Principle -> Impact -> Fix -> Verification -> Self-Review -> Confidence -> Limitations. Ground all statements strictly in provided evidence. | Vulnerability ID: GHSA-g66v-54v9-52pr
Source: osv
Title: Vikunja has SSRF via Todoist/Trello Migration File Attachment URLs that Allows Reading Internal Network Resources
Description: ## Summary
The migration helper functions `DownloadFile` and `DownloadFileWithHeaders` in `pkg/modules/migration/helpers.go` make arbit... | STRUCTURED SECURITY REASONING: GHSA-g66v-54v9-52pr
============================================================
-> UNDERSTAND:
Target package 'code.vikunja.io/api' (Go) is affected by Server-Side Request Forgery (SSRF) (CWE-918).
Advisory summary: Vikunja has SSRF via Todoist/Trello Migration File Attachment URLs that... | null | null | null | null | null |
CYBER_M_FINAL_052 | Memory safety | Advanced | 88 | Verified | Rank vulnerability findings using a defensive risk prioritization matrix based on exploitability, impact, and complexity. | Vulnerability ID: GHSA-g66v-54v9-52pr
Source: osv
Title: Vikunja has SSRF via Todoist/Trello Migration File Attachment URLs that Allows Reading Internal Network Resources
Description: ## Summary
The migration helper functions `DownloadFile` and `DownloadFileWithHeaders` in `pkg/modules/migration/helpers.go` make arbit... | DEFENSIVE RISK PRIORITIZATION MATRIX: GHSA-g66v-54v9-52pr
============================================================
ASSESSED SEVERITY: CVSS_V3
ACTION RANK: P3 (LOW PRIORITY)
1. EXPLOITABILITY FACTORS:
- Attack Vector: Network / Untrusted Input
- Authentication Required: Dependent on environment
2. IMPACT AS... | null | null | null | null | null |
CYBER_M_FINAL_053 | Memory safety | Advanced | 98 | Verified | Identify the trust boundaries relevant to this vulnerability. Which boundary was violated, what controls should exist at each boundary, and how should they be reinforced? | Vulnerability ID: GHSA-g66v-54v9-52pr
Source: osv
Title: Vikunja has SSRF via Todoist/Trello Migration File Attachment URLs that Allows Reading Internal Network Resources
Description: ## Summary
The migration helper functions `DownloadFile` and `DownloadFileWithHeaders` in `pkg/modules/migration/helpers.go` make arbit... | ### Technical Security Analysis
TRUST BOUNDARY ANALYSIS: GHSA-g66v-54v9-52pr
============================================================
Trust boundaries are points where data transitions between different
trust levels. Vulnerabilities occur when data crosses a boundary
without proper validation.
SYSTEM TRUST BOUND... | null | null | null | null | null |
End of preview. Expand in Data Studio
🛡️ Cybersecurity Dataset V4 (Cyber_M Master Release)
Cybersecurity Dataset V4 is a curated, de-boilerplated, and Abstract Syntax Tree (AST)-verified dataset engineered specifically for fine-tuning enterprise defensive cybersecurity language models. It represents the official training and evaluation corpus powering CyberQwen-14B-Security-V2.
The dataset features 1,668 active master records partitioned into four strictly disjoint splits with 0% cross-split code or prompt leakage.
🔍 Data Provenance & Sources: Where the Data Came From
Cybersecurity Dataset V4 was compiled from multi-source vulnerability databases, open-source vulnerability fix commits, compiler diagnostics, and production telemetry:
- Open Source Vulnerabilities (OSV.dev) & GitHub Security Advisories (GHSA):
- Real-world vulnerability records, vulnerability advisories, and commit-level before-and-after patch diffs across open-source ecosystems (PyPI, npm, Go, Maven, Cargo, Packagist, RubyGems).
- National Vulnerability Database (NVD / NIST) & MITRE CWE:
- Authoritative vulnerability descriptions, Common Weakness Enumeration (CWE) hierarchical taxonomy relationships, and Common Attack Pattern Enumeration and Classification (CAPEC) vectors.
- Compiler Sanitizer Crash Dumps (ASan / MSan / UBSan):
- Crash traces and diagnostic output from AddressSanitizer, MemorySanitizer, and UndefinedBehaviorSanitizer derived from real-world C/C++ memory corruption vulnerabilities (heap buffer overflows, use-after-free, uninitialized reads).
- CWE-Bench & Real-World Open-Source Repositories:
- 100 enterprise and open-source software repositories spanning 91 distinct CWE classes across 8 languages, curated to provide ground-truth remediation pairs.
- Static Analysis Rule Schemas (Semgrep & AST Matchers):
- Schema-validated Semgrep rules (
rules: [ { id, message, severity, languages, pattern } ]) verified directly against the nativesemgrep --validateengine.
- Schema-validated Semgrep rules (
- SOC & Threat Forensics Telemetry:
- Synthesized and sanitized Windows Security Event Logs (Event ID 4688 Process Creation, Event ID 4624 Logon), Linux Auditd/Sysmon logs, and Zeek network session records mapping directly to MITRE ATT&CK enterprise tactics.
⚙️ Data Curation & Quality Engineering Pipeline
The V4 release underwent a multi-stage forensic curation process:
- De-Boilerplating & Information Density Optimization:
- Removal of generic conversational preamble, uninformative docstrings, and synthetic filler to maximize technical information density per token.
- AST Sink Inspection & Remediation Verification:
- Every code repair example was parsed to confirm that vulnerable AST sinks (e.g.,
eval(),shell=True, unparameterized SQL concatenation, path traversal sinks, insecure deserialization) were completely eradicated and replaced with defensive patterns.
- Every code repair example was parsed to confirm that vulnerable AST sinks (e.g.,
- Cryptographic Deduplication:
- Exact, near-duplicate, and semantic duplicate filtering using normalized cryptographic SHA-256 hashing across code snippets and task instructions.
- Zero Cross-Split Leakage Guarantee:
- Rigorous pairwise audit across all 7 split pairs (Train vs. Val, Train vs. Test, Train vs. Benchmark, Val vs. Test, etc.) verifying 0 code collisions and 0 prompt collisions (100% clean isolation).
📊 Dataset Structure & Split Breakdown
The master dataset contains 1,668 records partitioned into four disjoint splits:
| Partition | File Name | Record Count | % of Master | Description & Role |
|---|---|---|---|---|
| Train | train.jsonl |
1,251 | 75.0% | Supervised fine-tuning split for CyberQwen-14B-Security-V2. |
| Validation | validation.jsonl |
166 | 10.0% | Model checkpointing and hyperparameter tuning. |
| Test | test.jsonl |
198 | 11.9% | Unseen offline evaluation holdout. |
| Benchmark | benchmark.jsonl |
53 | 3.2% | Fresh, independent holdout benchmark scenarios. |
| Master | master.jsonl |
1,668 | 100.0% | Complete master collection. |
🏷️ Category & Language Distribution
Security Categories:
- Authentication and Authorization: 138 records (OAuth2, JWT, IDOR, BFLA, MFA, session fixation)
- Injection Vulnerabilities: 125 records (SQLi, Command Injection, NoSQL, SSTI, LDAP/XPath)
- Multi-file Repository Reasoning: 125 records (Gateway router decouple, model-serializer leak, worker context)
- AI-Generated Insecure Code Detection: 125 records (ReDoS, insecure deserialization, prompt injection sinks)
- API and Business-Logic Security: 105 records (Negative quantities, coupon race conditions, mass assignment)
- Cloud, Container and Infrastructure Security: 95 records (IMDS SSRF
169.254.169.254, K8s RBAC, Docker root) - Cryptography and Session Management: 85 records (Static IV, weak PRNG, MD5/SHA1 collision, disabled TLS)
- Path Traversal and File Handling: 75 records (Zip Slip, canonicalization bypass, unrestricted upload)
- Dependency and Supply-Chain Security: 65 records (Dependency confusion, unpinned versions, postinstall hooks)
- False-Positive and Negative-Control Examples: 55 records (Parameterized SQL, DOMPurify, safe subprocess)
- Incomplete Evidence and Uncertainty: 35 records (Missing router context, unknown sanitizers, ambiguous flags)
- Secure Patch and Regression Testing: 45 records (Incomplete replace bypass, regex flaws, regression tests)
- Baseline Historical Categories: 479 records (Root-cause analysis, memory safety, concurrency, incident telemetry)
Programming Languages Covered:
- Python: 261 records
- TypeScript: 129 records
- Go: 110 records
- Java: 105 records
- JavaScript / Node.js: 100 records
- C# / .NET: 58 records
- PHP: 58 records
- Ruby: 58 records
- Rust: 58 records
- C / C++: 58 records
- JSON / npm / pip manifests: 65 records
- Baseline Multi-language: 608 records
📋 Data Schema
Each record contains the following normalized fields:
{
"id": "CYBER_M_V4_0042",
"category": "Injection vulnerabilities",
"difficulty": "hard",
"quality_score": 0.98,
"evidence_status": "verified_ast",
"instruction": "Identify the vulnerability in the following Go HTTP handler and refactor it securely:",
"input": "func handler(w http.ResponseWriter, r *http.Request) {\n target := r.URL.Query().Get(\"target\")\n cmd := exec.Command(\"ping\", \"-c\", \"1\", target)\n out, _ := cmd.CombinedOutput()\n w.Write(out)\n}",
"output": "The handler is vulnerable to CWE-78 (OS Command Injection)...",
"vulnerability_class": "CWE-78",
"synthetic": false,
"reasoning_skill": "AST sink inspection & secure refactoring",
"verification_method": "Static Code & Semantic Analysis",
"metadata": {
"cwe": "CWE-78",
"language": "Go",
"provenance": "Cyber_M_V4_High_Depth_Expansion",
"task_type": "Vulnerability remediation",
"vulnerability_status": "confirmed_vulnerability"
}
}
🚀 Associated Models
- CyberQwen-14B-Security-V2: Flagship cybersecurity model trained on V4, achieving 88.0% on the Enterprise Defensive Benchmark, 99.0% patch rate on CWE-Bench, and 99.6% on ExploitGym.
- CyberQwen Interactive ZeroGPU Demo: Live interactive playground.
⚖️ Intended Use & Ethics
This dataset is released exclusively for defensive cybersecurity research, automated vulnerability mitigation, DevSecOps pipeline triage, and static analysis development. It does NOT contain weaponized exploit payloads or dual-use malware automation.
- Downloads last month
- 5
Models trained or fine-tuned on AarthShah/cybersecurity-dataset-v4