OFFENSIVE SECURITY
Vulnerability research · Exploitation fundamentals · Stack overflows · Modern mitigations · ROP · Heap exploitation · Web/network exploits · Fuzzing
Vulnerability Research Methodology
- CVSS v3.1 scores vulnerabilities 0-10 across Base, Temporal, and Environmental metric groups.
- CVE identifies specific vulnerabilities; CWE classifies the underlying weakness type.
- Responsible disclosure gives vendors time to patch before public details are released.
- Bug bounty platforms (HackerOne, Bugcrowd, Intigriti) let researchers earn rewards legally.
- Patch diffing reveals exactly what a vendor fixed, which often points directly to the vulnerability.
The security industry uses standardised systems to track, score, and communicate vulnerabilities. These systems let defenders prioritise what to patch first and let researchers publish findings in a structured way.
Walk into Naivas looking for cooking oil. "Cooking oil" is a whole shelf category — that is a CWE, the kind of weakness. But the 2-litre Elianto bottle with batch number 4471 sitting on that shelf is one exact product — that is a CVE, this specific flaw in this specific software. The shop tracks both the category and the exact item on the shelf, and so do security researchers.
► CVE and CWE - The Classification Systems
A CVE (Common Vulnerabilities and Exposures) is a specific instance of a vulnerability in a specific product. A CWE (Common Weakness Enumeration) is the category of weakness that caused it. The distinction matters: CWE-79 is "Cross-site Scripting" (the weakness type); CVE-2021-44228 is Log4Shell (the specific instance in Log4j).
| TERM | WHAT IT IS | EXAMPLE |
|---|---|---|
| CVE | Unique ID for a specific vulnerability in a specific product | CVE-2021-44228 (Log4Shell in Log4j) |
| CWE | Category of weakness that leads to vulnerabilities | CWE-79 XSS, CWE-89 SQLi, CWE-119 Buffer Overflow |
| NVD | NIST National Vulnerability Database - enriches CVEs with CVSS scores | nvd.nist.gov |
| MITRE | Maintains the CVE list and CWE list | cve.mitre.org, cwe.mitre.org |
| Exploit-DB | Public exploit archive, searchable by CVE or product | exploit-db.com |
| Rapid7 VulnDB | Commercial vuln database with deep technical detail | vulndb.rapid7.com |
► CVSS Scoring - How Severity is Measured
CVSS (Common Vulnerability Scoring System) produces a 0.0-10.0 score. Version 3.1 is current; version 4.0 was released in 2023. The score has three metric groups:
| GROUP | METRICS | WHAT IT MEASURES |
|---|---|---|
| Base | Attack Vector, Attack Complexity, Privileges Required, User Interaction, Scope, Confidentiality/Integrity/Availability Impact | Intrinsic severity of the flaw itself, unchanged by time or environment |
| Temporal | Exploit Code Maturity, Remediation Level, Report Confidence | How the severity changes over time (PoC available? Patch released?) |
| Environmental | Modified Base metrics + Confidentiality/Integrity/Availability Requirements | Severity adjusted for YOUR specific environment and asset criticality |
| CVSS RANGE | SEVERITY | TYPICAL MEANING |
|---|---|---|
| 0.1 - 3.9 | LOW | Difficult to exploit, minimal impact |
| 4.0 - 6.9 | MEDIUM | Some exploitation difficulty or limited impact |
| 7.0 - 8.9 | HIGH | Relatively easy to exploit with significant impact |
| 9.0 - 10.0 | CRITICAL | Network-exploitable, no auth, full system compromise possible |
► Vulnerability Lifecycle
Understanding where a vulnerability sits in its lifecycle determines how urgent patching is and what exploit options exist:
DISCOVERY Researcher or attacker finds the flaw
|
v
PROOF OF CONCEPT Minimal code confirming exploitability (PoC)
|
v
WEAPONISATION PoC refined into reliable exploit (may include evasion)
|
v
PATCH RELEASE Vendor releases fix
|
v
N-DAY WINDOW Vulnerable systems still unpatched - race against time
|
v
PATCH COMPLETE Exposure minimised (never zero - some never patch)► Zero-Day vs N-Day
A zero-day (0-day) is a vulnerability that has no available patch - "zero days" for defenders to patch before attackers exploit it. An N-day is a vulnerability with a patch available for N days. N-days are often more dangerous in practice because they are well-documented and widely weaponised, but many organisations patch slowly.
► Disclosure Models
| MODEL | APPROACH | PROS | CONS |
|---|---|---|---|
| Responsible/Coordinated | Researcher notifies vendor privately, gives fix deadline (typically 90 days per Google Project Zero), then publishes | Gives vendors time to patch; industry standard | Vendors may drag feet; stressful for researcher |
| Full Disclosure | Researcher publishes all details immediately with no vendor notification | Forces vendors to act; helps defenders too | Gives attackers immediate weaponisable details |
| Bug Bounty (HackerOne, Bugcrowd, Intigriti) | Researcher submits via platform, receives monetary reward per programme scope | Legal, paid, structured, often includes coordinated disclosure | Programme scope limits what you can test; triage delays |
► Patch Diffing
When a vendor releases a security patch, the difference between the patched and unpatched binary reveals exactly what was fixed. Tools like BinDiff (Zynamics/Google) and Diaphora (open source) perform binary diffing to identify changed functions. Researchers use this to reconstruct the vulnerability from the patch - often faster than independent discovery.
1. Download vulnerable version binary 2. Apply vendor patch to get patched binary 3. Load both into BinDiff/Diaphora 4. Identify changed functions (highlighted by diff tool) 5. Analyse changed function in Ghidra/IDA to understand the fix 6. Work backwards to understand the exploitable path in the old version
- CVSS v3.1 scores vulnerabilities 0-10 across Base, Temporal, and Environmental metric groups.
- CVE identifies specific vulnerabilities; CWE classifies the underlying weakness type.
- Responsible disclosure gives vendors time to patch before public details are released.
- Bug bounty platforms (HackerOne, Bugcrowd, Intigriti) let researchers earn rewards legally.
- Patch diffing reveals exactly what a vendor fixed, which often points directly to the vulnerability.