Prev Next

Tools / Common Vulnerability Attack Types & Recent Activity Interview questions

Last updated

1. What is a vulnerability in cybersecurity? 2. What is the difference between a vulnerability, a threat and a risk? 3. What is a CVE? 4. What is CVSS? 5. What is the CISA KEV catalog? 6. What is a zero-day vulnerability? 7. What is SQL injection? 8. What is cross-site scripting (XSS)? 9. What is CSRF? 10. What is remote code execution (RCE)? 11. What is a buffer overflow? 12. What is command injection? 13. What is path traversal? 14. What is broken access control (IDOR)? 15. What is SSRF? 16. What is XXE? 17. What is privilege escalation? 18. What is a phishing attack? 19. What is ransomware? 20. What is a DDoS attack? 21. What is a man-in-the-middle attack? 22. What is credential stuffing? 23. What is a software supply chain attack? 24. What is the OWASP Top 10? 25. What is a race condition (TOCTOU)? 26. What is the difference between CVE and CWE? 27. What is the difference between stored, reflected and DOM XSS? 28. How do parameterized queries stop SQL injection? 29. Why is a CVSS score alone not enough to prioritize patching? 30. How does CSRF protection with tokens and SameSite work? 31. What is the difference between SSRF and CSRF? 32. Why are VPNs and edge devices attacked so often? 33. How do attackers chain vulnerabilities together? 34. What happens when an unrestricted file upload flaw is exploited? 35. How did the ChainDrop npm worm spread? 36. How can a CI pipeline be abused to publish trusted malicious packages? 37. Why was the Axios npm compromise so serious? 38. What is the difference between typosquatting and dependency confusion? 39. Why is insecure deserialization dangerous? 40. How does prompt injection differ from SQL injection? 41. How do SSRF attacks reach cloud metadata services? 42. Explain the execution flow of a ransomware attack? 43. How do you troubleshoot a suspected web shell? 44. How does HSTS stop SSL stripping? 45. How can you optimize patching for KEV-listed flaws? 46. When should you choose a WAF over fixing the code? 47. How can you defend against credential stuffing? 48. What is the difference between a vulnerability scan and a penetration test? 49. How do you mitigate DDoS attacks? 50. Which recently exploited vulnerabilities should candidates know about?

1. What is a vulnerability in cybersecurity?

A vulnerability is a weakness in software, hardware, configuration or process that an attacker can use to harm confidentiality, integrity or availability.

The weakness itself does no damage. It becomes a problem when someone can reach it and exploit it. Common forms:

  • Code flaw - unchecked input, memory errors, logic bugs.
  • Design flaw - an admin API that never asks who is calling.
  • Misconfiguration - default passwords, open storage buckets, debug mode left on.

Interviewers often follow up with how flaws get tracked (CVE) and scored (CVSS).

Take quiz
A vulnerability by itself is best described as:
a weakness that turns dangerous once it can be reached and exploited
an attack that is already underway
a firewall rule that blocks suspicious traffic
a log entry recording a failed login
Which is a misconfiguration-type vulnerability?
a phishing email sent to finance staff
a default admin password left unchanged on a router
a flood of requests from a botnet
a laptop stolen from a car

2. What is the difference between a vulnerability, a threat and a risk?

These three words are not interchangeable, and mixing them up is a classic interview slip.

Term Meaning Example
Vulnerability A weakness that can be exploited Unpatched VPN appliance
Threat An actor or event that could exploit it A ransomware group scanning for that VPN
Risk Likelihood of exploitation combined with business impact Chance of a breach times the cost of downtime and data loss

Remove any one of the three and the risk drops. Patching removes the vulnerability, while controls such as network segmentation reduce the impact.

Take quiz
An unpatched VPN appliance is an example of a:
threat actor
risk register entry
vulnerability
compensating control
Risk is best expressed as:
the total number of CVEs on a server
the age of the operating system
the length of the patch window
likelihood of exploitation combined with business impact

3. What is a CVE?

CVE stands for Common Vulnerabilities and Exposures. It is a public catalog that gives each disclosed security flaw a unique ID, such as CVE-2026-46300.

IDs are issued by CVE Numbering Authorities (CNAs), including vendors like Microsoft and Cisco, under the program run by MITRE. The ID only names the flaw so that scanners, advisories and patch notes all talk about the same thing.

It does not carry a severity score. That comes separately from CVSS, usually published through the NVD or by the vendor.

Take quiz
What does a CVE ID tell you by itself?
a unique identifier for one publicly disclosed flaw
the exact CVSS severity score
whether a patch exists
which threat group exploits it
In CVE-2026-46300, the year 2026 generally reflects:
the CVSS version used for scoring
the year the ID was assigned or reserved
the year the vendor shipped the patch
the number of affected products

4. What is CVSS?

CVSS (Common Vulnerability Scoring System) rates the technical severity of a flaw from 0.0 to 10.0. It has three metric groups: Base (intrinsic traits), Temporal/Threat (exploit maturity, available fixes) and Environmental (your own setup).

Score Severity
0.0 None
0.1 - 3.9 Low
4.0 - 6.9 Medium
7.0 - 8.9 High
9.0 - 10.0 Critical

CVSS v4.0 is the newest version, though v3.1 scores still appear on most advisories. Remember that CVSS measures severity, not the likelihood that anyone will attack you.

Take quiz
A CVSS base score of 9.4 falls in which band?
High
Medium
Critical
Low
Which metric group reflects details of your own deployment?
Base
Temporal/Threat
Vendor advisory
Environmental

5. What is the CISA KEV catalog?

The Known Exploited Vulnerabilities (KEV) catalog is CISA's list of CVEs with reliable evidence of active exploitation in the wild.

US federal civilian agencies must fix KEV entries by a due date set under a binding operational directive (recent CISA notices cite BOD 26-04). Everyone else treats the list as a priority signal: if a flaw is on KEV, someone is already using it.

Recent additions show the pattern: a WordPress core race condition, Ivanti Connect Secure flaws, and a Brotli library flaw all landed in October and November 2026. Deadlines can be as short as three days.

Take quiz
A CVE is added to KEV when:
there is evidence it is being exploited in the wild
its CVSS score passes 9.0
the vendor ships a patch
a researcher publishes a blog post about it
Binding KEV deadlines apply directly to:
every private company
Federal Civilian Executive Branch agencies
only cloud providers
only open-source maintainers

6. What is a zero-day vulnerability?

A zero-day is a flaw that attackers exploit before the vendor has a patch, so defenders have had zero days to prepare.

A recent example: researchers at Volexity reported in October 2026 that Ivanti Connect Secure flaws were being used as zero-days in a chain that ended in code execution.

Since no patch exists, you lean on compensating controls: remove internet exposure where possible, segment the device, watch for odd behavior, and apply vendor mitigations. Once the fix ships, the flaw is no longer a zero-day, though it stays dangerous until patched.

Take quiz
What makes a flaw a zero-day?
it has a CVSS score of 10
it only affects operating systems
attackers exploit it before a patch is available
it was found on January 1
Best response while no patch exists:
wait quietly for the vendor and change nothing
turn off logging to reduce noise
lower the CVSS score in the scanner
restrict exposure, segment, monitor and apply vendor mitigations

7. What is SQL injection?

SQL injection happens when untrusted input becomes part of a SQL statement, so the attacker changes what the query does.

The root cause is string concatenation:

query = "SELECT * FROM users WHERE name = '" + name + "'"
# input:  ' OR '1'='1
# result: the WHERE clause is always true

Impact ranges from data theft to login bypass and, on weak setups, data destruction. The fix is parameterized queries, plus least-privilege database accounts.

Take quiz
The root cause of SQL injection is:
user input concatenated into a query string
weak password hashing
a missing HTTPS certificate
oversized cookies
Best primary fix:
hiding error pages only
parameterized queries (prepared statements)
blocking the word SELECT in the browser
longer session timeouts

8. What is cross-site scripting (XSS)?

XSS lets an attacker run their own script in a victim's browser under a trusted site's origin. The script can read page data, steal session tokens that lack HttpOnly, or act as the user.

It appears when an app puts untrusted data into a page without proper encoding. The three flavors are stored, reflected and DOM-based (covered later).

  • Output encoding for the right context (HTML, attribute, JS).
  • Content-Security-Policy to limit what scripts can run.
  • HttpOnly cookies so scripts cannot read them.
Take quiz
XSS executes attacker script in:
the database engine
the server's operating system shell
the victim's browser, under the trusted site's origin
the DNS resolver
Which header helps limit XSS damage?
Content-Length
Accept-Language
X-Request-ID
Content-Security-Policy

9. What is CSRF?

Cross-Site Request Forgery tricks a logged-in user's browser into sending a state-changing request the user never intended, such as changing an email address or transferring funds.

It works because browsers automatically attach cookies to requests for the site. The attacker cannot read the response, so CSRF targets actions, not data.

Defenses: anti-CSRF tokens, SameSite cookies, and checking the Origin header on sensitive requests.

Take quiz
CSRF works mainly because:
browsers automatically send cookies with requests to the site
passwords are stored in plain text
the server runs HTTP/2
script gets injected into the page
A solid CSRF defense is:
gzip compression
anti-CSRF tokens combined with SameSite cookies
longer URLs
removing the Secure flag from cookies

10. What is remote code execution (RCE)?

RCE lets an attacker run their own code or commands on a target system from afar, usually with the privileges of the vulnerable service.

That is why it sits at the top of most severity rankings: it can lead to data theft, persistence and lateral movement. Typical causes are command injection, insecure deserialization, memory corruption and unsafe file upload.

Recent KEV entries show the pattern: the Ivanti Connect Secure chain ended in code execution, and the PHP2Shell flaws were used to drop web shells.

Take quiz
Why is RCE ranked among the most severe flaw types?
it only reveals software version numbers
it only crashes the user interface
the attacker can execute code with the service's privileges
it only affects the attacker's own machine
A common RCE root cause is:
a slow database index
oversized image files
dark mode CSS
insecure deserialization of untrusted data

11. What is a buffer overflow?

A buffer overflow occurs when a program writes more data into a fixed-size memory block than it can hold, spilling into adjacent memory.

It mostly affects C and C++ code. Results range from a crash to corrupted data to attacker-controlled code execution, depending on what gets overwritten.

Defense What it does
Bounds checking Rejects writes past the buffer end
Memory-safe languages Rust, Go, Java prevent most overflows by design
ASLR Randomizes memory layout
Stack canaries / NX Detect smashed stacks, block code running from data pages

Take quiz
A buffer overflow happens when a program:
writes more data than the buffer was sized for
reads from a closed file
opens too many browser tabs
uses too much disk space
Which defense randomizes memory layout?
CSRF tokens
ASLR
HSTS
password salting

12. What is command injection?

Command injection occurs when an app passes user input to an operating system shell, letting the attacker append their own commands.

# Vulnerable
os.system("ping -c 1 " + host)

# Safer: no shell, arguments as a list
subprocess.run(["ping", "-c", "1", host], check=True)

Beyond avoiding the shell, validate input against an allowlist (a hostname pattern, for example) and run the process with minimal privileges.

Take quiz
Best way to avoid command injection:
strip only the space character
hide the endpoint URL
call the program with an argument list, skip the shell and validate input
base64 encode the host name
Unlike SQL injection, command injection sends input to:
a database engine
a browser DOM
a TLS handshake
an operating system shell

13. What is path traversal?

Path traversal (directory traversal) abuses ../ sequences in a file path parameter to read or write files outside the intended folder.

For example, a download endpoint that builds a path from ?file=report.pdf might be tricked into serving a system file instead.

The reliable fix is to canonicalize the final path and confirm it still sits inside the allowed base directory. Allowlisting file names and avoiding user-controlled paths altogether is better still.

Take quiz
Path traversal usually abuses:
../ sequences in a file path input
expired TLS certificates
duplicate cookies
oversized JSON arrays
A strong defense is to:
rename the folder weekly
resolve the canonical path and confirm it is inside the base directory
block only the word 'etc'
limit file sizes

14. What is broken access control (IDOR)?

Broken access control means the app fails to enforce what a user is allowed to do. The best-known form is IDOR (insecure direct object reference), where changing an identifier exposes someone else's data.

Example: a logged-in user edits /invoices/1042 to /invoices/1043 and sees another customer's invoice.

The fix is a server-side ownership check on every request. Unpredictable IDs help a little but do not replace authorization. It has held the #1 spot in the OWASP Top 10 through the 2025 edition.

Take quiz
The core fix for IDOR is:
hiding the ID in a hidden form field
encrypting the URL in the browser
a server-side check that the user owns the requested object
adding a CAPTCHA
In the OWASP Top 10:2025, broken access control ranks:
last
tenth
removed from the list
first (A01)

15. What is SSRF?

Server-Side Request Forgery lets an attacker make the vulnerable server send requests to destinations the attacker picks, often internal systems the attacker cannot reach directly.

A typical trigger is a feature that fetches a URL for you: webhooks, image import, link previews. Attackers point it at internal admin panels or cloud metadata endpoints.

Defenses: allowlist destinations, block private and link-local ranges after DNS resolution, disable unneeded redirects, and require the hardened metadata service on cloud hosts.

Take quiz
In SSRF, who sends the request to the internal target?
the vulnerable server itself
the victim's browser
the attacker's home router
a DNS root server
SSRF commonly targets:
the user's clipboard
internal services and cloud metadata endpoints
a local printer queue
CPU microcode

16. What is XXE?

XML External Entity injection occurs when an XML parser resolves external entity references from untrusted input. That can expose local files, trigger SSRF or cause denial of service.

It needs three ingredients: untrusted XML, a parser with entity or DTD processing enabled, and a way for the result to reach the attacker.

Recent activity: CISA added a Cisco BroadWorks XXE flaw (CVE-2026-20320) to KEV on November 10, 2026. The fix is to disable DTDs and external entity resolution in the parser.

Take quiz
XXE is caused by:
a JSON parser rejecting data
a CSS selector bug
an XML parser resolving external entities from untrusted XML
an expired JWT
Primary fix:
convert XML to uppercase
compress the XML payload
raise the XML size limit
disable DTD and external entity processing in the parser

17. What is privilege escalation?

Privilege escalation means gaining more access than you were granted. It usually follows an initial foothold.

Type Meaning Example
Vertical Normal user becomes admin/root Local kernel bug gives root
Horizontal Access another user's data at the same level Reading a coworker's files

CISA noted that the Linux 'Fragnesia' flaw (CVE-2026-46300) can be chained with PHP2Shell bugs to implant web shells and run commands as root. The main control is least privilege plus prompt kernel patching.

Take quiz
Moving from a normal user to root is:
vertical privilege escalation
horizontal privilege escalation
session fixation
DNS tunneling
Which principle limits the damage of a compromised account?
security through obscurity
least privilege
default allow
shared admin credentials

18. What is a phishing attack?

Phishing uses deceptive messages to trick people into giving up credentials, running malware or approving payments.

Variant Channel / target
Spear phishing Crafted for one specific person
Whaling Senior executives
Smishing SMS text messages
Vishing Phone calls

Filters and training help, but phishing-resistant MFA (FIDO2 keys or passkeys) is the strongest fix because it will not work on a fake domain.

Take quiz
Phishing crafted for one specific person is called:
smishing
vishing
spear phishing
pharming
Which MFA method resists phishing best?
SMS one-time codes
security questions
emailed codes
FIDO2 security keys or passkeys

19. What is ransomware?

Ransomware is malware that encrypts a victim's files, or locks systems, and demands payment for recovery.

Most modern gangs use double extortion: they steal data first, then threaten to publish it, so restoring from backup alone does not end the pressure.

Practical defenses: patch internet-facing systems quickly (edge devices are a favorite way in), enforce MFA, segment networks, and keep offline or immutable, tested backups.

Take quiz
Double extortion means:
encrypting files and also threatening to leak stolen data
asking for payment in two currencies
infecting two operating systems at once
encrypting every backup twice
The most reliable recovery safeguard is:
paying the ransom quickly
offline or immutable backups that are regularly tested
renaming encrypted files
turning off antivirus

20. What is a DDoS attack?

A Distributed Denial of Service attack floods a target with traffic from many sources, usually a botnet, so legitimate users cannot get through.

Type Target Example
Volumetric Bandwidth UDP reflection/amplification
Protocol Connection tables SYN flood
Application (L7) App resources HTTP request flood on a search page

Mitigation combines upstream scrubbing or a CDN, rate limiting and autoscaling. Layer 7 floods are the hardest to spot because each request looks legitimate.

Take quiz
What makes a DoS attack 'distributed'?
one very large server sends it
it only uses DNS traffic
traffic arrives from many sources, often a botnet
it always uses encryption
An HTTP request flood against a search page is:
a physical-layer attack
a data-link attack
a UDP reflection attack
an application-layer (L7) attack

21. What is a man-in-the-middle attack?

In a man-in-the-middle (MITM) attack, someone secretly sits between two parties, reading or altering their traffic.

Common setups: a rogue Wi-Fi hotspot, ARP spoofing on a local network, DNS hijacking, or SSL stripping that downgrades HTTPS to HTTP.

TLS defends against it by encrypting traffic and authenticating the server's certificate. Add HSTS, certificate validation in clients, and avoid unknown public Wi-Fi without a VPN.

Take quiz
TLS helps against MITM by:
encrypting traffic and authenticating the server certificate
hiding your IP address
compressing responses
blocking ICMP ping
A rogue Wi-Fi hotspot that relays victim traffic is an example of:
SQL injection
a man-in-the-middle attack
a buffer overflow
XXE

22. What is credential stuffing?

Credential stuffing is the automated replay of username and password pairs leaked from one breach against other sites, hoping people reused them.

It differs from brute force: attackers use real, previously valid credentials, so each guess has a decent chance. Tools rotate IPs to dodge simple rate limits.

Defenses covered later: MFA, breached-password checks, bot detection and sensible rate limiting.

Take quiz
Credential stuffing relies on:
trying every possible character combination
a kernel buffer overflow
people reusing passwords across sites
an unpatched database
It differs from brute force because it uses:
random character strings
hardware tokens
stolen TLS certificates
real leaked credential pairs

23. What is a software supply chain attack?

A supply chain attack compromises something you trust and depend on, such as a library, build pipeline or update channel, so malicious code reaches many downstream users at once.

2026 has been busy. The Axios npm package was compromised on March 31 to deliver a remote access trojan, and the self-spreading ChainDrop worm hit 400+ npm packages on August 4. Both follow the Shai-Hulud pattern that began in September 2025.

OWASP's 2025 list now has Software Supply Chain Failures as its own category (A03).

Take quiz
A supply chain attack compromises:
a trusted upstream component or build process used by many
only the target's firewall
only users' passwords
only physical servers
Shai-Hulud-style worms mainly spread by:
exploiting kernel drivers
stealing publish tokens and republishing infected packages
scanning for open RDP ports
phishing customers' browsers

24. What is the OWASP Top 10?

The OWASP Top 10 is a community-driven awareness list of the most critical web application security risks. It is not a testing checklist, and each entry bundles many CWEs.

The 2025 edition (published November 2025) keeps broken access control at #1. Security misconfiguration is #2, and Software Supply Chain Failures is a new #3. Injection dropped to #5, and Mishandling of Exceptional Conditions is new at #10. SSRF is now folded into access control.

OWASP also maintains a separate Top 10 for LLM applications, where prompt injection is first.

Take quiz
Which category sits at #1 in OWASP Top 10:2025?
Injection
Cryptographic Failures
Broken Access Control
Insecure Design
Which is a new category in the 2025 list?
Cross-Site Scripting
Sensitive Data Exposure
Using Components with Known Vulnerabilities
Software Supply Chain Failures

25. What is a race condition (TOCTOU)?

A race condition exists when the outcome depends on the timing of events. TOCTOU (time-of-check to time-of-use) is a specific case: the state changes between the check and the action that relies on it.

if os.access(path, os.W_OK):   # check
    f = open(path, 'w')          # use - path may now point elsewhere

CISA added a WordPress core TOCTOU flaw (CVE-2026-119977) to KEV on November 3, 2026, with active exploitation confirmed. The fix pattern is to make check and use one atomic operation or hold a lock.

Take quiz
TOCTOU happens when:
state changes between a check and its use
two users pick the same password
a cookie expires before login
a cache is flushed at midnight
The standard fix is to:
add a delay between them
make the check and the use a single atomic operation
run the check twice
log the check result

26. What is the difference between CVE and CWE?

A CVE identifies one specific flaw in a specific product, while a CWE (Common Weakness Enumeration) names the general class of mistake behind it.

CVE CWE
One real flaw in a real product version A type of weakness, e.g. CWE-89 SQL injection
Example: CVE-2026-48356 (Magento file upload) Example: CWE-434 unrestricted upload of dangerous file type
Used for patching and tracking Used for secure coding and root-cause trends

Many CVEs map to the same CWE. That is why the OWASP Top 10 categories are built from CWE groups rather than individual CVEs.

Take quiz
CWE-89 (SQL injection) is best described as:
a single flaw in one product
a severity score
a weakness class
a vendor patch ID
Which is used mainly for tracking one specific flaw across products and patches?
CWE
OWASP category
CVSS vector
CVE

27. What is the difference between stored, reflected and DOM XSS?

All three run attacker script in the victim's browser. They differ in where the malicious data lives and which side processes it.

Type Where the payload lives Who triggers it
Stored Saved in the database (comments, profiles) Any visitor who loads the page
Reflected In the request, echoed in the response A victim who clicks a crafted link
DOM-based Handled only by client-side JavaScript A victim, with no server involvement in the flaw

Stored XSS usually has the widest blast radius. For DOM XSS, review how client code uses sinks like innerHTML, and prefer textContent.

Take quiz
Which XSS type is saved on the server and hits every visitor?
stored XSS
reflected XSS
DOM-based XSS
self-XSS only
DOM-based XSS is unusual because the flaw is in:
the database schema
client-side JavaScript handling data
the TLS configuration
the load balancer

28. How do parameterized queries stop SQL injection?

A parameterized query separates code from data. The database receives the SQL structure first and parses it, then receives the values separately and treats them strictly as data.

cur.execute("SELECT * FROM users WHERE name = %s", (name,))

Even if name contains quote characters or SQL keywords, it can never change the parsed statement. Escaping tries to clean the input; parameterization removes the ambiguity.

Limits: you cannot parameterize table or column names, so allowlist those. Pair this with a low-privilege DB account.

Take quiz
Why can't injected text change a parameterized query?
the database encrypts the input
the input is converted to uppercase
the statement is parsed first and values are bound strictly as data
the driver deletes quote characters
Which part of a query cannot be bound as a parameter?
string values
numeric values
date values
table or column names

29. Why is a CVSS score alone not enough to prioritize patching?

CVSS measures how bad a flaw could be, not how likely it is to be used against you. A 9.8 that nobody exploits may matter less than a 7.5 that is under active attack.

Better prioritization layers more signals:

  1. KEV listing - confirmed exploitation in the wild.
  2. EPSS - a probability score for exploitation in the next 30 days.
  3. Exposure - is it internet-facing?
  4. Asset value - what does the system hold or connect to?

Only 26% of detected KEV flaws were fully remediated in one 2026 study, which suggests many teams still struggle to rank work well.

Take quiz
Which signal confirms a flaw is being attacked now?
a KEV catalog listing
a high CVSS base score
a long CVE description
the vendor's logo on the advisory
EPSS estimates:
the patch download size
the probability a flaw will be exploited soon
the CVSS vector string
the vendor's response time

30. How does CSRF protection with tokens and SameSite work?

The two controls attack different halves of the problem.

An anti-CSRF token is a random, per-session (or per-request) value embedded in forms or sent in a header. The server rejects state-changing requests without the right token. A cross-site attacker cannot read it because of the same-origin policy.

SameSite=Lax or Strict tells the browser not to attach the cookie on most cross-site requests, so the forged request arrives without the session.

sequenceDiagram
  participant A as Attacker page
  participant B as Browser
  participant S as Bank server
  A->>B: Hidden form posts to bank
  B->>S: Request (SameSite blocks cookie)
  S-->>B: 403 - no session / token

Use both: SameSite is a strong default, tokens cover older browsers and same-site edge cases.

Take quiz
Why can't an attacker's page read the anti-CSRF token?
tokens are always encrypted with RSA
browsers hide all hidden form fields
the same-origin policy blocks cross-site reads
tokens expire in one second
What does SameSite=Strict do?
encrypts the cookie value
extends the cookie lifetime
forces HTTP/3
withholds the cookie on cross-site requests

31. What is the difference between SSRF and CSRF?

The names are similar but the victims and mechanics differ.

SSRF CSRF
Who sends the forged request The server The user's browser
Abuses trust in The server's network position The browser's saved session cookie
Typical goal Reach internal services or cloud metadata Perform an action as the logged-in user
Main defense Destination allowlists, network egress controls Tokens and SameSite cookies

In short, SSRF tricks a server into making a call, while CSRF tricks a user's browser into making one.

Take quiz
In CSRF, the forged request is sent by:
the victim's browser
the application server
the database
the CDN origin shield
A defense that fits SSRF but not CSRF is:
SameSite cookies
an outbound destination allowlist
anti-CSRF tokens
an Origin header check

32. Why are VPNs and edge devices attacked so often?

Edge devices (VPN gateways, firewalls, load balancers, file-transfer tools) sit on the internet, hold powerful credentials and often cannot run an EDR agent. One flaw can give a foothold inside the network.

Attackers also favor them because patching is slow: updates need downtime, and devices are often forgotten in asset inventories.

Recent KEV additions in October and November 2026 included Ivanti Connect Secure flaws exploited as zero-days, and a reported OpenSSH issue. Mitigations: minimize exposure, patch on KEV timelines, forward device logs to a SIEM, and use integrity checkers when vendors offer them.

Take quiz
Why does compromising an edge device pay off for attackers?
it only stores marketing files
it always runs the newest patches
it sits on the internet and often holds privileged access into the network
it has no network connectivity
Which control helps most when an edge device lacks an EDR agent?
disabling its logging
adding more users to it
exposing its admin page publicly
forwarding its logs to a SIEM and watching for anomalies

33. How do attackers chain vulnerabilities together?

Attackers often combine several mid-severity flaws, because each one provides a step the next one needs.

Reported examples: the Ivanti Connect Secure exploit chain paired a remote authenticated file upload flaw with a separate authentication bypass leading to code execution. CISA also said Fragnesia (CVE-2026-46300) can be chained with the PHP2Shell flaws to plant a web shell and run commands as root.

flowchart LR
  A["Auth bypass"] --> B["File upload"]
  B --> C["Web shell"]
  C --> D["Privilege escalation"]
  D --> E["Root access"]

Defender takeaway: do not dismiss a 'medium' flaw if it sits next to others on the same asset.

Take quiz
Why can chaining make 'medium' flaws dangerous?
each flaw supplies a step the next one needs
medium flaws are always zero-days
chaining lowers the CVSS score
chains only work offline
CISA said Fragnesia can be chained with PHP2Shell to:
encrypt DNS traffic
implant web shells and run commands as root
disable BGP routing
patch itself

34. What happens when an unrestricted file upload flaw is exploited?

The attacker uploads a file the server should have rejected, often a script, and then requests it so the server executes it. That turns an upload form into remote code execution with a web shell as the usual result.

CISA added an Adobe Commerce and Magento unrestricted upload flaw (CVE-2026-48356) to KEV, and the Ivanti chain also used an upload bug.

  1. Validate type by content, not only by extension or MIME header.
  2. Store uploads outside the web root, or on a separate domain.
  3. Rename files and strip execute permissions.
  4. Scan uploads and cap size.
Take quiz
Why is unrestricted upload so dangerous?
it only wastes disk space
it only changes the page color
the server may execute the uploaded file, giving code execution
it only affects email
Which is a good storage practice for uploads?
inside the web root with original names
in the cookie
in the URL
outside the web root with execute permission removed

35. How did the ChainDrop npm worm spread?

ChainDrop, a Shai-Hulud-family worm, hit npm on August 4, 2026 and compromised over 400 packages within hours.

  1. A trojanized package ran a preinstall hook on a developer machine or CI runner.
  2. It hunted for npm publish tokens and GitHub tokens.
  3. With a valid token it pulled every package the account could publish, added the malicious hook, bumped the patch version and republished.
  4. Stolen secrets went to attacker-controlled GitHub repositories.
flowchart LR
  A["Install infected package"] --> B["Steal tokens"]
  B --> C["Republish maintainer packages"]
  C --> D["New victims install"]
  D --> A

Unit 42 also saw the operator change command-and-control settings through a single Ethereum transaction, with no malware update needed.

Take quiz
What did the worm steal to keep spreading?
npm publish and GitHub tokens
browser bookmarks
BIOS passwords
Wi-Fi SSIDs
Which npm feature did it abuse to run on install?
package-lock integrity checks
the preinstall lifecycle hook
semantic version ranges only
the README file

36. How can a CI pipeline be abused to publish trusted malicious packages?

The May 2026 TanStack incident showed that even legitimate release workflows can be turned against a project.

Reported steps: attackers poisoned a CI cache, the worm scraped tokens from the GitHub Actions runner's memory, and then exchanged them for npm publish credentials through npm's own OIDC token exchange. No npm token was 'stolen' in the traditional sense.

The result was malicious packages carrying valid SLSA provenance, so provenance proved where the build happened, not that the build was clean.

  • Do not share caches across trust boundaries (PR builds vs release builds).
  • Scope OIDC trust to specific workflows and branches.
  • Require approval for release jobs.
  • Monitor for unexpected publishes.
Take quiz
Why did provenance not stop the TanStack attack?
provenance is never generated
attestations are always forged offline
the malicious build ran inside the legitimate pipeline
npm does not support provenance
Which hardening step addresses cache poisoning?
longer package names
bigger runners
disabling README files
isolating caches between untrusted PR builds and release builds

37. Why was the Axios npm compromise so serious?

On March 31, 2026, a malicious Axios release appeared on npm. Axios has over 100 million weekly downloads, so the blast radius was huge.

Researchers reported that the bad version pulled in a package that had not existed before that day, which dropped a remote access trojan for Windows, Linux and macOS. That is classic installer malware.

Lessons: pin and review dependency updates, add a short cooldown before adopting brand-new releases, use lockfiles, and watch for new transitive dependencies appearing in a minor update.

Take quiz
What made the Axios compromise high impact?
the package has a huge install base across many ecosystems
it only affected one company
it was a hardware bug
it needed physical access
Which habit would have reduced exposure to a brand-new bad release?
always using 'latest' tags
a cooldown period before adopting newly published versions
disabling lockfiles
installing as root

38. What is the difference between typosquatting and dependency confusion?

Both trick a build into pulling a malicious package, but through different mistakes.

Typosquatting Dependency confusion
Trick A look-alike name, like reqeusts A public package using the same name as a private one
Who errs A developer typing the name The package manager picking the higher public version
Fix Review names, use allowlists Scoped names, private registry priority, reserved names

A related 2026 pattern: campaigns impersonating brands, such as fake Asurion-named packages found by Panther.

Take quiz
Dependency confusion happens when:
a developer misspells a name
a CDN goes offline
a public package shadows a private package name
a lockfile is too short
A look-alike package name like 'reqeusts' is:
dependency confusion
XXE
clickjacking
typosquatting

39. Why is insecure deserialization dangerous?

Deserialization rebuilds objects from bytes. If those bytes come from an attacker, the process can trigger code paths the developer never intended, such as 'gadget chains' in existing classes.

Result: remote code execution, authentication bypass or denial of service, depending on the language and libraries loaded.

  • Never deserialize untrusted data with native formats (Java serialization, Python pickle).
  • Prefer simple formats like JSON with schema validation.
  • Sign or encrypt serialized blobs and verify before parsing.
  • Keep libraries updated.
Take quiz
What is a 'gadget chain'?
existing classes combined to run unintended code during deserialization
a series of CSS rules
a chain of firewall rules
a Git branching strategy
Which is safer for untrusted input?
Python pickle
JSON with schema validation
native Java serialization
eval of the string

40. How does prompt injection differ from SQL injection?

Both mix untrusted input with trusted instructions, but SQL has a clear boundary between code and data. LLMs do not.

SQL injection is fixed by parameterization. With a prompt, instructions and data are both plain text, so no equivalent clean separation exists.

The risk grows with agents: a model reading a malicious web page or repo file may be tricked into running tools. Reports in 2026 described AI coding agents bypassed through shell tricks, and a worm poisoning AI assistants through MCP server injection.

  • Treat model output as untrusted.
  • Give agents least-privilege tools and require approval for risky actions.
  • Isolate untrusted content from instructions where possible.
Take quiz
Why is prompt injection harder to fix than SQL injection?
LLMs cannot read text
SQL databases are slower
instructions and data are both plain text with no hard boundary
prompts are always encrypted
A sound control for an AI agent reading untrusted content is:
giving it admin keys
disabling all logging
trusting its output blindly
least-privilege tools and approval for risky actions

41. How do SSRF attacks reach cloud metadata services?

Cloud instances expose a metadata endpoint on a link-local address, 169.254.169.254, that can hand out instance details and temporary credentials.

If an app fetches attacker-supplied URLs, the attacker points it at that address, and the app returns the credentials. Those credentials then work from anywhere until they expire.

  • Require the session-token style metadata service (IMDSv2 on AWS).
  • Block link-local and private ranges after DNS resolution.
  • Give instance roles minimal permissions.
  • Allowlist outbound destinations.
Take quiz
Why is 169.254.169.254 a popular SSRF target?
it serves instance metadata and temporary credentials
it hosts public DNS
it is the default web port
it runs the CDN
Which control makes metadata theft harder?
using longer hostnames
requiring session tokens for the metadata service
enabling directory listing
disabling TLS

42. Explain the execution flow of a ransomware attack?

Most intrusions follow the same arc, and defenders get chances to break it at every stage.

flowchart LR
  A["Initial access"] --> B["Credential theft"]
  B --> C["Lateral movement"]
  C --> D["Data exfiltration"]
  D --> E["Backup destruction"]
  E --> F["Encryption + ransom note"]
  1. Initial access - phishing, stolen credentials or an exposed edge device flaw.
  2. Discovery and privilege escalation - find domain admin.
  3. Exfiltration - copy data for double extortion.
  4. Impact - disable backups, then encrypt at scale, often out of hours.

Best detection points: unusual admin logins, mass file reads, and security tool tampering before encryption starts.

Take quiz
What usually happens just before mass encryption?
the attacker announces themselves
DNS is turned off permanently
backups are deleted or disabled and data is exfiltrated
all passwords are reset by IT
A good detection point earlier in the chain is:
a slow web page
a font change
a printer jam
unusual admin logins and lateral movement

43. How do you troubleshoot a suspected web shell?

Treat it as a live incident. Evidence first, cleanup second.

  1. Preserve evidence - snapshot the host and copy web and auth logs.
  2. Find the file - look for recently changed scripts in web directories, odd names and obfuscated code.
  3. Check the logs - repeated POSTs to one file, strange user agents, requests spiking at night.
  4. Find the entry point - match upload or exploit requests to the first appearance.
  5. Contain - isolate the host, rotate secrets the web app could read.
  6. Rebuild and patch - redeploy from known-good source.

Do not just delete the file. If the entry flaw is still open, it will come back.

Take quiz
What should you do first on a suspected web shell?
preserve evidence such as snapshots and logs
delete the file immediately
reboot everything
post about it publicly
Why not just delete the web shell?
deleting is illegal
the underlying entry flaw may still be open
web shells are harmless
it will delete itself

44. How does HSTS stop SSL stripping?

In SSL stripping, a MITM intercepts the first plain HTTP request and keeps the user on HTTP while talking HTTPS to the real server.

HSTS (HTTP Strict Transport Security) tells the browser, via the Strict-Transport-Security header, to use HTTPS only for the domain for a set time. After the first visit, the browser upgrades requests itself and refuses certificate errors.

Remaining gap: the very first visit. The HSTS preload list closes it by baking the domain into browsers.

Take quiz
What does HSTS tell the browser?
cache images longer
block all cookies
use HTTPS only for this domain for a set time
use HTTP/1.0
What gap does HSTS preload close?
weak passwords
slow DNS
server-side injection
the unprotected first visit

45. How can you optimize patching for KEV-listed flaws?

Treat KEV as your top tier and build a process that can hit its short deadlines.

  1. Ingest the feed - match KEV entries to your asset inventory automatically.
  2. Rank by exposure - internet-facing assets first.
  3. Pre-approve emergency changes so a 3-day deadline does not wait on a weekly board.
  4. Mitigate first - if you cannot patch, restrict access or apply vendor workarounds.
  5. Verify - rescan; for compromised-risk flaws, also hunt for signs of earlier exploitation.

Remember the Ivanti guidance: patching alone may not be enough, and CISA asked for forensic triage as well.

Take quiz
What is the first step in a KEV-driven process?
match KEV entries to your asset inventory
wait for the monthly patch day
reboot all servers
lower CVSS thresholds
If you cannot patch by the deadline, you should:
ignore the entry
apply mitigations such as restricting exposure
delete the asset record
disable monitoring

46. When should you choose a WAF over fixing the code?

A web application firewall is a compensating control, not a substitute for fixing the bug. Choose it when you need time or cannot change the code.

Situation Choose
Vendor product with no patch yet WAF rule as virtual patch
Legacy app nobody can safely change WAF plus planned rewrite
Your own code with a known injection flaw Fix the code, WAF as a backup
Layer 7 DDoS or bot traffic WAF / bot management

WAFs can be bypassed with encoding tricks, so rules need tuning and the root cause still needs closing.

Take quiz
A WAF is best described as:
a replacement for secure coding
a database index
a compensating control that can virtually patch while a fix is prepared
a password manager
For your own code with an injection bug, you should:
only add a WAF rule
remove input validation
hide the form
fix the code and keep the WAF as a backup

47. How can you defend against credential stuffing?

No single control stops it, so layer several.

  • MFA - the strongest control, since a reused password alone no longer works.
  • Breached-password checks - reject known leaked passwords at signup and reset.
  • Rate limiting and bot detection - throttle per account and per IP range, and challenge suspicious flows.
  • Anomaly alerts - flag logins from new devices or impossible travel.
  • Passkeys - remove the password altogether.

Per-IP limits alone fail because attackers rotate through huge proxy pools, so limit per account as well.

Take quiz
Which control most directly neutralizes a reused leaked password?
multi-factor authentication
longer cookie lifetimes
bigger password boxes
hiding the login URL
Why is per-IP rate limiting alone weak?
IP addresses are encrypted
attackers rotate through large proxy pools
rate limits break HTTPS
browsers ignore them

48. What is the difference between a vulnerability scan and a penetration test?

A vulnerability scan is automated and broad: it checks systems against a database of known flaws and misconfigurations. A penetration test is a goal-driven, mostly manual engagement that tries to exploit weaknesses, chain them and show real impact.

Vulnerability scan Penetration test
Method Automated Human-led, tool-assisted
Output List of potential issues Proven attack paths and impact
Frequency Continuous / weekly Annual or after major change
False positives More common Fewer, since findings are validated

They complement each other. Scans keep coverage wide while pen tests answer 'how bad could it really get?'

Take quiz
Which activity validates and chains flaws to show impact?
a vulnerability scan
a log rotation job
a penetration test
an asset tag audit
A vulnerability scan is mainly:
a manual red team campaign
a legal audit
a code review of one function
automated, broad detection of known issues

49. How do you mitigate DDoS attacks?

Mitigation is layered, because different attacks hit different resources.

  1. Absorb - put a CDN or DDoS scrubbing service in front so volumetric traffic never reaches your origin.
  2. Filter - block spoofed and malformed packets at the edge, use SYN cookies.
  3. Rate limit - throttle expensive endpoints like search and login.
  4. Scale - autoscale stateless tiers and cache aggressively.
  5. Plan - have a runbook and contacts at your ISP and provider.

Hide the origin IP, or attackers will go around the CDN.

Take quiz
Why hide the origin server's IP behind a CDN?
so attackers cannot bypass the protection and hit it directly
to speed up DNS
to hide the logo
to raise TTLs
Which control helps against SYN floods?
CSRF tokens
SYN cookies
HSTS preload
password salting

50. Which recently exploited vulnerabilities should candidates know about?

A quick snapshot of what CISA and researchers reported around October and November 2026. Check the KEV catalog for the latest, since additions land weekly.

Item Type Why it matters
Ivanti Connect Secure (CVE-2026-111357, -118527) Upload flaws chained with auth bypass Exploited as zero-days; forensic triage requested
PHP2Shell + Fragnesia (CVE-2026-46300) Web shell chain plus Linux privilege escalation Root-level command execution
WordPress core (CVE-2026-119977) TOCTOU race condition Added to KEV Nov 3
Adobe Commerce / Magento (CVE-2026-48356) Unrestricted file upload Classic path to web shells
Cisco BroadWorks (CVE-2026-20320) XXE Added to KEV Nov 10
ChainDrop (Aug 2026) npm worm, 400+ packages Token theft and automatic republishing

In an interview, name the class of flaw first, then the product. It shows you understand the pattern, not just the headline.

Take quiz
Which item in the list is an XXE flaw?
WordPress core CVE-2026-119977
Magento CVE-2026-48356
Cisco BroadWorks CVE-2026-20320
Fragnesia CVE-2026-46300
The WordPress core flaw added to KEV on Nov 3 is a:
buffer overflow
XSS flaw
SSRF flaw
TOCTOU race condition
«
»

Comments & Discussions