1/140
A comprehensive collection of vocabulary-style flashcards covering the core concepts, principles, attacks, and countermeasures of COM535 Cyber Security Weeks 1 through 8.
CIA Triad
The foundational security model consisting of Confidentiality (only authorised parties can read), Integrity (data is not altered undetected), and Availability (service is reachable when needed).
Confidentiality
A security property ensuring that only authorised parties can read data; broken by sniffing, Man-in-the-Middle (MitM) attacks, or data breaches.
Integrity
A security property ensuring that data cannot be altered undetected; broken by tampering or DNS poisoning.
Availability
A security property ensuring that a service remains reachable when needed; broken by Distributed Denial of Service (DDoS), ransomware, or hardware failure.
Vulnerability vs. Threat vs. Control
A vulnerability is a weakness; a threat is something that exploits a weakness; a control is a countermeasure that protects the weakness.
Phishing Varieties (Email vs. Spear vs. Whaling)
Email phishing is mass-distributed; spear phishing is targeted and personalised; whaling specifically targets executives. Common defences include training, filtering, MFA, and URL inspection.
Modern Ransomware Trends
Emerging trends including double extortion (stealing data and threatening to leak it), Ransomware-as-a-Service (RaaS), and the targeting of critical infrastructure.
Insider Threat Types
Three categories of insider risks: malicious (disgruntled employees), negligent (careless behaviour), and compromised (stolen credentials).
Zero-Day
A vulnerability that is unknown to the vendor and unpatched, leaving no defence until a security patch is released.
Supply Chain Attack
An attack that compromises a trusted supplier to compromise downstream victims. Notable examples include SolarWinds (2020), NotPetya (2017), and the npm Shai-hulud worm (2025).
Advanced Persistent Threat (APT)
A prolonged, targeted intrusion in which an attacker maintains undetected access over an extended period, typically to exfiltrate data.
Zero Trust (NIST SP 800-207)
A security model summarized as 'never trust, always verify', granting no implicit trust based on network location; it is identity-centric, relies on micro-segmentation, and enforces least privilege.
General Data Protection Regulation (GDPR)
A regulation requiring consent/transparency, data minimisation, data subject rights (access, rectify, erase), and a 72-hour breach notification window, backed by fines up to EUR 20m or 4% of global turnover.
Security Information and Event Management (SIEM)
A system providing centralised collection, correlation, and analysis of logs across an entire organization as a single pane of glass (e.g., Splunk, QRadar, Sentinel, Elastic).
IDS vs. IPS
An Intrusion Detection System (IDS) passively detects and alerts; an Intrusion Prevention System (IPS) sits inline to block, drop, or reset malicious connections in real time.
Snort Default Mode
The default operational execution mode of the Snort intrusion detection engine, which is passive.
PKI Certificate Revocation
A necessary mechanism because keys can be compromised or certificates forged; because issued certificates have long lifespans, they remain trusted until explicitly invalidated via a Certificate Revocation List (CRL).
IAAA Model
A security framework consisting of Identification (who you claim to be), Authentication (proving who you are), Authorisation (what actions you may perform), and Accountability (auditing what actions you took).
HTTP 401 vs. HTTP 403
HTTP 401 Unauthorized denotes an authentication failure (the server does not know who you are); HTTP 403 Forbidden denotes an authorization failure (the user is authenticated but not permitted access).
Authentication Factors
Five distinct authentication categories: something you know, something you are, something you have, location, and behavioural attributes.
FAR, FRR, and EER in Biometrics
FAR (False Acceptance Rate) is the rate an impostor is accepted (security failure); FRR (False Rejection Rate) is the rate a genuine user is rejected (usability failure); EER (Equal Error Rate) is the threshold where FAR=FRR (lower represents a better system).
Phishing-Resistant MFA
Authentication methods bound cryptographically to the real site origin, exclusively consisting of FIDO2 keys and passkeys. SMS, email, TOTP apps, and push notifications are susceptible to relay and phishing.
Rainbow-Table Attack
An attack utilizing precomputed hash chains for rapid password lookup; effectively mitigated by appending a unique, random salt to each user's password prior to hashing.
Credential Stuffing vs. Password Spraying
Credential stuffing reuses breached username/password pairs; password spraying tests a single common password across many accounts to avoid lockout thresholds. Both are mitigated by MFA.
Secure Password Storage
Storing passwords using a one-way cryptographic hash, a unique salt (to defeat rainbow tables), and key stretching (to slow down brute-force attempts). Recommended OWASP order: Argon2id > scrypt > bcrypt > PBKDF2.
NIST SP 800-63B Password Guidelines
Guidelines advising at least 15+ characters if used as a sole factor, allowing lengths of 64+, eliminating composition rules and forced periodic expiry, screening against breach lists, and forbidding hints or security questions.
Passkeys / FIDO2
A phishing- and breach-resistant credential standard where the site stores only a public key (no shared secret); the browser automatically supplies the true origin, rendering signatures gathered by phishing sites invalid.
User Enumeration
A vulnerability in which an application reveals whether a username exists through error messaging or response timing; mitigated with generic responses, consistent response timing, and rate limiting.
Saltzer & Schroeder Design Principles
Fundamental security design principles: least privilege, fail-safe defaults, complete mediation, separation of privilege, economy of mechanism, open design, and psychological acceptability.
Reference Monitor Properties
Three mandatory properties of an access-control reference monitor: it must be always invoked (tamper-resistant/cannot be bypassed), tamperproof, and verifiable (compact enough to analyze thoroughly).
Access Control List (ACL) vs. Capability List
An ACL associates access permissions with the object ('who can access this file?'); a capability list associates permissions with the subject ('what can Bob access?'), behaving like a delegable ticket that is difficult to revoke.
DAC vs. MAC vs. RBAC vs. ABAC
DAC allows the object owner to set access (e.g., ACLs); MAC enforces central system policy via mandatory security labels; RBAC assigns access based on organizational roles; ABAC dynamically evaluates attributes of user, resource, action, and context.
Bell-LaPadula vs. Biba Models
Dual formal access control models: Bell-LaPadula enforces confidentiality (no read up, no write down); Biba enforces integrity (no read down, no write up).
Symmetric vs. Asymmetric vs. Hashing
Symmetric cryptography uses one shared key (e.g., AES, fast bulk encryption); asymmetric uses a public/private key pair (e.g., RSA/ECDH, for signatures and key exchange); hashing requires no key and provides one-way integrity verification.
Kerckhoffs's Principle
A foundational cryptosystem principle stating that a system's security must depend solely on the secrecy of the cryptographic key, not on maintaining the secrecy of the algorithm.
Data Encryption Standard (DES) Demise
DES failed because its 56-bit key and 64-bit block size were too small (brute-forced in 1998); it was temporarily succeeded by Triple-DES and ultimately replaced by AES (an SPN cipher featuring a 128-bit block and 128/192/256-bit keys).
Electronic Codebook (ECB) Mode
A block cipher mode that encrypts each block independently, producing identical ciphertext for identical plaintext blocks, which leaks structural patterns (such as image outlines).
RSA Security Foundation
An asymmetric cryptosystem based on the mathematical difficulty of factoring large composite integers n=p×q; theoretically broken by Shor's algorithm on a quantum computer.
Diffie-Hellman Key Exchange
A method allowing two parties to establish a shared secret over an insecure channel based on the discrete logarithm problem; plain Diffie-Hellman lacks authentication and is vulnerable to MitM attacks unless paired with digital certificates.
Cryptographic Hash Properties
The core guarantees of secure hash functions: preimage resistance (one-way), second-preimage resistance (cannot find a duplicate input matching a given hash), collision resistance (cannot find any two matching inputs), and the avalanche effect.
Message Authentication Code (MAC) vs. Digital Signature
A MAC relies on a shared symmetric key and cannot provide non-repudiation; a digital signature utilizes a private key, ensuring non-repudiation because only the private key holder could have generated it.
Digital Certificate
A credential that cryptographically binds a public key to an identity via a Certificate Authority (CA) signature, establishing a chain of trust from the server certificate through intermediate CAs to a trusted root CA.
TLS 1.3 Architecture
A secure protocol combining ECDHE key exchange (providing forward secrecy), certificate signatures (preventing MitM), HKDF key derivation, and an AEAD cipher (AES-GCM or ChaCha20) for combined confidentiality and integrity.
Harvest Now, Decrypt Later
A post-quantum threat where adversaries intercept and store encrypted traffic today to decrypt once quantum computing matures; mitigated using AES-256 and NIST post-quantum cryptography standards (ML-KEM, ML-DSA).
Buffer Overflow
An anomaly occurring when a program writes more data into a fixed-length block of allocated memory than it can hold, potentially corrupting control flow.
gets() Vulnerability
A dangerous legacy C function that performs no buffer length checks, enabling excessive input to overflow memory and overwrite return addresses; safely replaced by fgets(), which enforces a maximum buffer size.
Stack-Smashing Attack
An exploitation technique that overflows a stack buffer to overwrite the saved return address, redirecting program execution to malicious attacker-controlled shellcode.
Shellcode
Attacker-supplied machine code engineered to perform specific actions (such as spawning a root shell); typically position-independent, architecture-specific, and automated by frameworks like Metasploit.
Unsafe C Functions and Replacements
Standard unsafe functions and their secure counterparts: gets() replaced by fgets(), strcpy() replaced by strlcpy(), strcat() replaced by strlcat(), and sprintf() replaced by snprintf().
Use-After-Free (UAF)
A memory corruption flaw caused by accessing a pointer after its allocated object has been freed, enabling an attacker to re-allocate that memory region with crafted malicious data.
TOCTOU Race Condition
Time-of-Check to Time-of-Use race condition occurring when system state alters between a security verification and resource access (e.g., swapping a file for a symlink); addressed using file handles and atomic operations.
Buffer Overflow Defences
Mitigation mechanisms against buffer overflows, including stack canaries, DEP/NX bit, ASLR, shadow stacks (CET), Control Flow Integrity (CFI), and memory-safe programming languages.
Stack Canary
A pseudo-random value written to the stack right before the saved return address; it is validated prior to function return, terminating the program immediately if modified.
Return-Oriented Programming (ROP)
An exploitation technique used to bypass NX/DEP by chaining existing instruction sequences ending in 'ret' (gadgets) within executable memory to execute attacker logic.
SQL Injection (SQLi)
An injection flaw where untrusted user input is directly concatenated into a SQL string, altering query structure (e.g., ' OR '1'='1); resolved by using parameterized queries and prepared statements.
Cross-Site Scripting (XSS)
A web vulnerability where malicious scripts execute within a victim's browser as trusted application content, often stealing session cookies; defended using context-aware output encoding, input validation, and CSP.
XSS Variants (Stored vs. Reflected vs. DOM-Based)
Stored XSS persists malicious input on the server (e.g., in database comments); reflected XSS immediately bounces unvalidated input off the server in a response/URL; DOM-based XSS executes purely within client-side script manipulation.
OS Command Injection
A flaw where user input containing shell metacharacters (; | &&) is passed unsanitised into an operating system shell, executing attacker commands; prevented by avoiding shell calls and using strict allow-listing.
Cross-Site Request Forgery (CSRF)
An attack forcing an authenticated user's browser to submit unauthorized state-changing requests using automatically included session cookies; mitigated by anti-CSRF tokens and SameSite cookie attributes.
Clickjacking
A client-side UI attack where an attacker conceals a target website inside an invisible iframe overlaying a malicious page, tricking users into clicking buttons; mitigated via X-Frame-Options or CSP frame-ancestors.
File-Upload Vulnerability
A weakness wherein an application accepts unverified files (such as shell.php) that can subsequently be requested to execute arbitrary server code; prevented by validating contents, renaming uploads, storing files outside the web root, and disabling execution permissions.
Open Redirect
An issue where a web application accepts untrusted URL parameters to direct user redirection, frequently leveraged in phishing campaigns; mitigated by allow-listing permitted target domains.
Server-Side Request Forgery (SSRF)
An attack forcing the backend server to make outbound requests to unintended destinations (e.g., internal networks or cloud metadata at 169.254.169.254); mitigated by allow-listing outbound URLs and blocking loopback/metadata IPs.
Allow-List vs. Deny-List Validation
Allow-listing accepts known-good inputs and is highly resilient; deny-listing blocks known-bad inputs and is fragile due to encoding evasion techniques. Input should always be canonicalised prior to validation.
Fuzzing
An automated testing technique providing semi-random, mutated input to programs to detect crashes or memory leaks; coverage-guided fuzzers (e.g., AFL, libFuzzer) optimize execution paths, and Google OSS-Fuzz applies it continuously.
SAST vs. DAST vs. IAST
Static Application Security Testing (SAST) evaluates source code without running it; Dynamic Application Security Testing (DAST) tests a running application externally; Interactive Application Security Testing (IAST) instruments the code internally during execution.
Virus vs. Worm vs. Trojan
A virus replicates by infecting other executable programs; a worm self-propagates automatically across network connections; a trojan disguises itself as benign utility while delivering a concealed malicious payload.
CWE Top 3 Weaknesses (2025)
The top three identified Common Weakness Enumerations: #1 XSS (CWE-79), #2 SQL injection (CWE-89), and #3 CSRF (CWE-352).
SPF, DKIM, and DMARC
Email security protocols: SPF validates authorized sending IPs; DKIM provides a cryptographic signature over message contents; DMARC verifies alignment with the visible From: domain and enforces handling policies and reporting.
Malvertising
The distribution of malicious code via legitimate online advertising networks onto trusted sites without requiring user clicks; defended by vetting ad networks, using CSP, patching browsers, and deploying ad-blockers.
Insufficient Logging
A systemic failure preventing early intrusion detection, incident investigations, and containment; mitigated by aggregating logs inside a SIEM, synchronizing system clocks, and establishing tamper-resistant retention.
Base-Rate Fallacy in IDS
A statistical phenomenon where, due to the extremely low base rate of true intrusions, an IDS with even 99% accuracy generates overwhelming volumes of false positives; tuning false-positive rates is vastly more impactful than headline accuracy.
DIRB
A web content exploration scanner that brute-forces hidden file and directory paths using wordlists, identifying unlinked web objects via HTTP status codes (such as 200 OK vs. 404 Not Found).
Nmap -sS (SYN Scan)
A stealthy TCP port scanning technique sending SYN packets without completing the 3-way handshake; a SYN/ACK response indicates an open port, RST indicates closed, and no reply indicates filtering.
Packet Filter vs. Stateful vs. Proxy Firewall
Packet filters inspect individual packet headers statelessly; stateful firewalls track active connection states in a state table; proxy firewalls (Application Layer Gateways) terminate and inspect the full Layer-7 session.
Firewall Default-Deny Rule
A security stance where all traffic not explicitly permitted by a rule is dropped, implemented by placing specific allow rules first and terminating with a broad block rule.
Firewall Limitations
Firewalls cannot prevent attacks encapsulated in legitimate traffic (e.g., phishing over HTTPS), stop insider threats, or monitor traffic that circumvents network perimeters (e.g., mobile hotspots, split tunnels).
SYN Flood Attack
A DoS attack transmitting spoofed SYN packets to overwhelm a server's connection backlog table with half-open connections; mitigated via SYN cookies, which encode session state into the sequence number without consuming memory.
Mirai Botnet
A massive 2016 DDoS botnet composed of approximately 600,000 compromised IoT devices using default Telnet credentials, which successfully disrupted Dyn DNS and critical web properties.
WEP vs. WPA2 vs. WPA3
Wireless security standards: WEP is broken (RC4, 24-bit IV); WPA2 uses AES-CCMP with a 4-way PSK handshake vulnerable to offline dictionary attacks; WPA3 introduces SAE (Dragonfly handshake), forward secrecy, and mandatory Protected Management Frames (PMF).
KRACK Attack
A 2017 vulnerability (Key Reinstallation Attack) that forces cryptographic nonce reuse in WPA2 by replaying handshake message 3, undermining stream encryption.
DNSSEC
A protocol suite providing origin authentication and data integrity for DNS records via cryptographic signatures; it does not encrypt DNS traffic (which requires DoT or DoH).
Link vs. End-to-End Encryption
Link encryption secures individual network hops but leaves payloads unencrypted at intermediate devices; end-to-end encryption preserves encryption across all intermediate nodes directly to the destination application (e.g., TLS).
Cyber Essentials Controls
The five baseline technical controls defined by the UK Cyber Essentials scheme: firewalls, secure configuration, security update management, user access control, and malware protection.
Trusted Computing Base (TCB)
The collective total of all hardware, firmware, and software components vital to enforcing a system's security policy; compromising any element inside the TCB breaks overall system security.
CPU Privilege Rings
Hardware-enforced security boundaries separating user applications (Ring 3) from the kernel and drivers (Ring 0), managed by a processor mode bit, with hypervisors functioning at Ring -1.
ASD Essential Eight Top Four
The Australian Signals Directorate's most critical controls: patching applications, patching operating systems, restricting administrative privileges, and application allow-listing, estimated to mitigate ≥85% of intrusions.
Secure Boot vs. Measured Boot
Secure Boot cryptographically verifies the signature of boot components before allowing them to execute; Measured Boot cryptographically hashes components into TPM Platform Configuration Registers (PCRs) to record a tamper-evident boot log.
Windows Security Reference Monitor (SRM)
The Windows kernel component that mediates object access by evaluating a subject's access token against the target object's security descriptor (ACLs) and generating security audit events.
Pass-the-Hash
An exploitation technique where an adversary uses a stolen NTLM password hash directly to authenticate against network services without cracking the cleartext password; mitigated by Credential Guard, LSA protection, and LAPS.
Linux Discretionary and Mandatory Controls
Standard Linux utilizes owner-driven rwx permissions (DAC) which can be augmented by Mandatory Access Control (MAC) systems such as SELinux (type enforcement labels) or AppArmor (path profiles); MAC can only further restrict access.
setuid-root Program Risk
A configuration wherein an executable runs with root privileges regardless of caller identity; any code flaw can lead directly to local privilege escalation (e.g., Dirty COW, PwnKit, Dirty Pipe, Baron Samedit).
xz Utils Backdoor (2024)
A sophisticated multi-year social-engineering infiltration that embedded an obfuscated backdoor into a core compression library linked to OpenSSH (sshd); highlights supply chain risks and the need for SBOMs and reproducible builds.
Virtual Machine vs. Container Isolation
A virtual machine uses a hypervisor to provide each tenant with an independent kernel (a 'wall'); containers share the host operating system kernel via namespaces/cgroups (a 'fence'), meaning a single kernel vulnerability can compromise all containers.
Prompt Injection
An LLM security vulnerability characterized as 'SQL injection without parameterized queries', where untrusted user input is mixed directly into system prompt instructions with no formal grammar to separate control instructions from data.
AI and Security Intersections
The three domains where AI interacts with cybersecurity: AI for security (defensive applications), security from AI (countering offensive attacks), and security of AI (defending models against adversarial threats).
Direct vs. Indirect Prompt Injection
Direct prompt injection occurs when an attacker directly inputs prompts to override system directions; indirect prompt injection occurs when the attacker embeds instructions in third-party content (e.g., web pages, documents) that the AI retrieves and parses.
The Lethal Trifecta (AI)
The co-occurrence of three capabilities in an autonomous AI agent: access to private data, exposure to untrusted inputs, and the permission to communicate externally; defending requires severing at least one leg (typically external egress).
Prompt-Injection Defence Failure
The empirical finding that twelve major published prompt-filtering defences failed in over 70% of tests against adaptive adversaries, establishing that heuristic filters do not form a robust security boundary.
CaMeL Architecture
A defensive prompt architecture where a privileged planner LLM uses trusted prompts, a quarantined LLM processes untrusted data without tools, output variables hold capability labels, and an interpreter enforces execution policies in deterministic code.