Security Vulnerability
Overview
A security vulnerability is a flaw arising in the design, implementation, configuration, or operation of an information system that, when exploited by an attacker, can lead to breaches of confidentiality, integrity, and availability (CIA) — such as data leaks, service disruption, or privilege theft. Unlike ordinary software bugs, the defining characteristic of a vulnerability is its "exploitability," and because a single vulnerability can escalate into a large-scale supply chain incident, it is treated as a central axis of modern cybersecurity management.
Key Details
Components of a Vulnerability
For a vulnerability to become a threat, three elements generally combine. First, a flaw inside the system (the vulnerability itself); second, a path by which that flaw can be reached (the attack surface); and third, an attacker or tool that exploits it (an exploit). Removing any one of these greatly reduces the risk.
Classification Frameworks: CVE, CWE, CVSS
- CVE (Common Vulnerabilities and Exposures): A unique identifier assigned to an individual vulnerability, standardized in formats such as CVE-2024-XXXXX, which serves as the basis for communication between vendors and researchers.
- CWE (Common Weakness Enumeration): A framework that classifies the "types" of vulnerabilities, codifying root causes such as buffer overflows, SQL injection, authentication bypass, and race conditions.
- CVSS (Common Vulnerability Scoring System): A scale that quantifies severity with a score from 0.0 to 10.0; CVSS v4.0 was released in late 2023, further refining the assessment of attack prerequisites and subsequent impact.
- CISA KEV: A list of vulnerabilities that have actually been exploited, widely used to determine patch priorities.
Representative Vulnerability Types
1. Injection attacks: Problems in which untrusted input — such as SQL injection, command injection, or XSS — is interpreted as code or a query; these have long ranked at the top of the OWASP Top 10.
2. Memory safety flaws: Buffer overflows, use-after-free, integer overflows, and similar issues are frequent in the C/C++ family and readily lead to remote code execution (RCE).
3. Authentication and authorization flaws: Weak password policies, session fixation, inadequate access control (IDOR), MFA bypass, and so on.
4. Cryptographic misuse: Hardcoded keys, weak randomness, use of deprecated algorithms.
5. Configuration errors and insecure defaults: Exposed administrative consoles, unnecessary services enabled, default accounts left in place.
6. Supply chain vulnerabilities: Open-source dependencies, build pipelines, unsigned updates, and so on.
Vulnerability Management Lifecycle
Vulnerability management operates as a cyclical structure: ① asset identification → ② scanning and detection (SAST/DAST/SCA, fuzzing, penetration testing) → ③ risk assessment and prioritization → ④ patching and mitigation → ⑤ verification and prevention of recurrence. In practice, since not all vulnerabilities can be fixed simultaneously, risk-based prioritization combining internet exposure, data sensitivity, actual exploitation status (such as KEV listing), CVSS score, and business impact has become the standard. In addition, it is recommended to understand dependency composition through SBOM (Software Bill of Materials) and to encourage reports from external researchers through vulnerability disclosure policies and bug bounty programs.
Legal and Institutional Context
In Korea, amendments to the Information and Communications Network Act and the Telecommunications Business Act have strengthened protections for vulnerability discovery and reporting as well as obligations to report infringement incidents, and regular vulnerability assessments of major information and communication infrastructure are conducted (led by KISA). Abroad, the United States' KEV and cybersecurity executive orders, and the EU's NIS2 Directive and Cyber Resilience Act (CRA), set security requirements at the point of product shipment.
Latest Trends
The trends of 2024–2025 can be summarized in four main strands. First, the popularization of AI-assisted vulnerability discovery. With the emergence of large language model-based code analysis tools and autonomous fuzzing agents, the pace of vulnerability discovery has accelerated sharply, and as a result the number of CVEs published simultaneously has reached an all-time high. Second, an increase in zero-day attacks targeting edge and network devices and the virtualization layer, with vulnerabilities in VPNs, firewalls, and hypervisors becoming major initial intrusion paths for ransomware groups. Third, the concretization of supply chain security regulation, as SBOM submission and code signing are being incorporated into procurement requirements, and the need to support open-source maintainers has been raised repeatedly. Fourth, vulnerabilities in AI systems themselves have emerged as a new category. Prompt injection, training data poisoning, model theft, and excessive agent privileges are being organized in the form of the OWASP LLM Top 10, with ongoing attempts to integrate them into existing application security frameworks. Furthermore, as patch cycles shorten and automatic updates become commonplace, the trend is to evaluate "understanding exposed assets and rapid mitigation" rather than "patch speed" as the core capability.
Related Topics
- [[Zero-Day Attack]]
- [[CVE]]
- [[CVSS]]
- [[Ransomware]]
- [[Supply Chain Attack]]
- [[Patch Management]]
- [[OWASP Top 10]]
- [[Bug Bounty]]
- [[SBOM]]
- [[Incident Response]]