Security+ SY0-801 Domain 2: Threats, Vulnerabilities, and Attacks

Domain 2 accounts for 24 percent of SY0-801. V8 separates threat and vulnerability characteristics, actors, vectors, attack surfaces, malicious indicators, and artificial intelligence (AI) security into six objectives.

A useful mental model is:

actor → motivation → vector/source → vulnerability or attack surface → malicious activity → evidence

The exact sequence can vary, but keeping those roles separate prevents many exam mistakes.

1. Domain 2 map

2. Threat and vulnerability characteristics

Threat intelligence and context

Threat intelligence becomes useful when it changes a decision. Sources may include internal observations, vendors, government advisories, information-sharing communities, vulnerability databases, and open sources.

Ask:

• Is the source timely?
• Is it relevant to the organization's technology and geography?
• Does it describe capability, intent, infrastructure, tactics, or exploitation?
• Is the information corroborated?
• Can the organization act on it?

Vulnerability identifiers and prioritization

These signals answer different questions.

Severity is not the same as exploit likelihood. Neither replaces asset criticality, exposure, compensating controls, or business impact.

3. Threat actors and motivations

Threat actors differ in resources, access, patience, and goals.

Common motivations include:

• financial gain
• espionage
• disruption
• ideology
• revenge
• competition
• geopolitical objectives
• curiosity or recognition

An insider is defined by authorized or trusted access, not automatically by malicious intent. Insiders can act deliberately or accidentally.

An advanced persistent threat usually implies sustained capability, planning, resources, and long-term objectives. Do not reduce the label to “very skilled attacker.”

4. Threat vectors and sources

A vector is the path or mechanism used to reach the target. A source identifies where the threat originates or is delivered.

Messaging and content

Email, text messages, collaboration platforms, attachments, documents, images, and quick response (QR) codes can all carry a malicious request or lead a user to a hostile destination.

Quishing uses QR codes as the phishing path. The risk is not that QR codes are inherently malicious. The code hides or shortens the visible destination and encourages the user to move the interaction to another device or browser context.

Browsers, extensions, password managers, cookies, and session tokens

Browser extensions can read or modify content depending on their permissions. Password managers protect credentials, but a compromised vault, malicious extension, weak recovery process, or deceptive site can still create risk.

A stolen authenticated session token or cookie can let an attacker reuse a session without performing the original authentication again. The security question then becomes session protection and revocation, not simply password strength.

Remote access and removable media

Remote administration, virtual private networks (VPNs), remote desktop services, vendor connections, and removable media create legitimate access paths that attackers may abuse.

Look for:

• exposed services
• weak credentials
• stale accounts
• missing multifactor authentication
• excessive privileges
• unmanaged endpoints
• untrusted media

Supply chain and service providers

A supplier, software dependency, logistics provider, software-as-a-service (SaaS) platform, managed service provider, or update channel can extend the attack path beyond the organization's direct control.

A supplier compromise does not remove the customer's responsibility to manage access, monitoring, segmentation, and recovery.

Wireless and proximity sources

Bluetooth, radio frequency (RF), near-field communication (NFC), Wi-Fi, and other wireless technologies change the physical boundary of an attack.

The range and protocol matter. NFC is intentionally short range. Bluetooth and Wi-Fi can operate at greater distances depending on power, environment, and configuration.

5. Vulnerabilities and attack surfaces

An attack surface is the collection of reachable opportunities an attacker could use. A vulnerability is a weakness that may be exploited.

Unsupported, unpatched, and misconfigured systems

Unsupported software may no longer receive fixes. Unpatched software has known updates that have not been applied. Misconfiguration can expose services, data, credentials, or administrative capabilities even when software is fully current.

Race conditions

A race condition occurs when security depends on the timing or order of concurrent operations and the application does not synchronize them safely. The attack may exploit a gap between checking a condition and using the result.

Hardcoded secrets and stale credentials

Hardcoded passwords, tokens, or keys are difficult to rotate and can leak through source repositories, images, logs, or deployments.

Stale credentials remain valid after they are no longer needed. Disable and remove them through lifecycle processes, not only after an incident.

Rogue devices and shadow IT

A rogue device is unauthorized or impersonates an approved device. Shadow IT is technology used outside approved management, procurement, or security processes.

Both create visibility gaps. The response should restore inventory, ownership, access control, logging, patching, and policy enforcement.

Public repositories and object storage

A public code repository can expose credentials, architecture details, or proprietary code. Public object storage can expose sensitive data through overly broad policies or configuration.

Do not assume “cloud” means private by default. Effective policy and access settings determine exposure.

Identity providers and large language models (LLMs) as attack surfaces

Identity providers concentrate authentication and federation trust. Misconfiguration, token theft, weak recovery, or privileged compromise can affect many downstream services.

Large language model (LLM) applications create attack surfaces through prompts, retrieved data, tools, plugins, model inputs, outputs, and connected systems.

6. Indicators of malicious activity

Indicators support a conclusion. One clue rarely proves the whole attack.

Credential and authentication clues

Examples include:

• repeated failed logins
• password spraying across many accounts
• successful authentication from an unusual device
• impossible-travel events
• new recovery methods
• session reuse from unexpected networks
• multifactor authentication (MFA) fatigue or bypass patterns
• user enumeration behavior

An impossible-travel event is an alert based on geography and timing. It should be investigated with device, session, token, network, and account evidence rather than treated as automatic proof of compromise.

Fileless and living-off-the-land activity

Fileless attacks may rely on memory, scripts, interpreters, or legitimate system tools rather than a traditional executable dropped to disk.

Living-off-the-land techniques abuse trusted tools already present in the environment. The tool name alone is not malicious. Context, parent process, command line, user, timing, destination, and surrounding activity matter.

Application and network indicators

Watch for:

• injection patterns
• unexpected outbound connections
• unusual Domain Name System (DNS) queries
• web shells
• lateral movement
• privilege escalation
• sudden data transfer
• command-and-control behavior
• changes to security settings
• abnormal process trees

Deepfakes and social engineering

Deepfake audio, video, or images can strengthen impersonation and urgency. The defense is process as much as technology.

For high-impact requests, use an independent trusted channel and a known verification process. A familiar voice or face should not replace authorization.

7. Artificial intelligence threats and vulnerabilities

Objective 2.6 is dedicated to AI security.

Prompt injection

Prompt injection supplies instructions intended to override or manipulate an AI system's intended behavior. Direct injection comes from the user. Indirect injection can arrive through content the model retrieves or processes.

Treat untrusted content as data, not authority. Limit tool permissions and validate high-impact actions outside the model.

Poisoning and model manipulation

Data poisoning alters training, fine-tuning, or retrieval data to influence later behavior. Model manipulation targets model parameters, artifacts, or deployment behavior.

Integrity controls, provenance, protected pipelines, validation, and monitoring help reduce the risk.

Evasion

Evasion changes input so the system produces a desired incorrect result while the underlying malicious goal remains. Small changes can matter when the model's decision boundary is sensitive.

Hallucinations and explainability

A hallucination is a confident but unsupported or false output. The response is not to trust confident language. Validate important facts against authoritative data.

Explainability concerns whether people can understand enough about a decision to evaluate, challenge, or govern it. Some systems need more transparency because the consequence of error is high.

Bias, privacy, and ethics

Training data, design choices, measurement, and deployment context can create uneven outcomes. Privacy risk can arise when systems expose sensitive prompts, retrieved records, training data, or generated outputs.

Governance should define acceptable data, human oversight, retention, testing, and escalation.

Jailbreaking and excessive agency

Jailbreaking attempts to bypass safeguards. Tool-enabled systems add risk when the model can take actions such as sending messages, changing data, running code, or calling privileged application programming interfaces (APIs).

Use least privilege, allowlists, confirmation for high-impact actions, isolation, audit logs, and deterministic authorization outside the model.

Session hijacking and code execution

AI applications still depend on ordinary web and application security. Stolen session tokens can hijack authenticated use. Unsafe tool invocation, generated commands, or code execution can turn model output into a system compromise.

8. Common Domain 2 traps

Choosing severity when the question asks likelihood

CVSS, EPSS, known exploitation, exposure, and business context answer different parts of prioritization.

Calling the delivery method the vulnerability

A phishing email is a vector. A hardcoded secret is a vulnerability. Stolen credentials or data exfiltration may be the resulting activity.

Calling every suspicious AI output prompt injection

A hallucination can occur without an attacker. Poisoning affects data or model integrity. Evasion manipulates input to influence a decision. Prompt injection manipulates instructions or context.

9. Domain 2 review checklist

Confirm that you can:

• distinguish CVE, CVSS, EPSS, and known exploitation
• match threat actors to capability and motivation without overgeneralizing
• identify messaging, QR, browser, session-token, supply-chain, remote-access, wireless, and physical vectors
• explain hardcoded secrets, stale credentials, shadow IT, rogue devices, race conditions, public repositories, public object storage, and identity-provider risk
• recognize fileless activity, living-off-the-land behavior, deepfakes, MFA bypass, credential attacks, and application indicators
• separate prompt injection, poisoning, evasion, hallucination, bias, privacy, jailbreaking, session hijacking, and unsafe AI tool execution

10. Official and primary references

• CompTIA Security+ certification page
• CVE Program
• FIRST Exploit Prediction Scoring System
• CISA Known Exploited Vulnerabilities Catalog
• NIST AI 100-2e2025: Adversarial Machine Learning
• OWASP Top 10 for Large Language Model Applications
• MITRE ATT&CK