Join our Newsletter — 33% off our NHI Course
Home› FAQ› Authentication, Authorisation & Trust› What is the difference between Argon2, PBKDF2, bcrypt,…
Authentication, Authorisation & Trust

What is the difference between Argon2, PBKDF2, bcrypt, and scrypt for password storage?

← Back to all FAQ
By NHI Mgmt Group Editorial Team Updated October 31, 2026 Domain: Authentication, Authorisation & Trust

Argon2 and scrypt are memory hard designs that make parallel cracking more expensive by using significant RAM as well as CPU. PBKDF2 and bcrypt are adaptive functions that slow hashing mainly through repeated computation. In practice, memory hard designs offer stronger resistance to modern attack hardware, while PBKDF2 often remains the compatibility driven fallback.

How the four password hashing functions differ in practice

All four are password storage primitives, but they are designed with different trade-offs. Argon2 and scrypt deliberately consume memory, which makes large-scale GPU and ASIC cracking less efficient. bcrypt and PBKDF2 are older adaptive key derivation functions that mainly increase work through repeated computation. The practical difference is not just algorithm name, but how much resistance they give against modern cracking hardware.

For teams comparing options, the useful question is whether the function is meant to slow an offline attacker who has stolen a password database. That is the core problem these algorithms address, and the answer depends on whether the design forces the attacker to spend mostly CPU time or both CPU and RAM.

Why memory hardness matters for password storage

Memory hardness changes the economics of cracking. An attacker can parallelise many hashes across GPUs, but a function that needs significant RAM per guess reduces how many guesses can run at once and raises the cost of each guess. That is why Argon2 and scrypt are generally preferred when you are choosing a fresh password storage scheme today.

By contrast, PBKDF2 and bcrypt remain useful because they are widely implemented and easier to support in older systems, libraries, and compliance driven environments. Their main strength is tunable work factor, but their resistance to highly parallel hardware is weaker than a well tuned memory hard design.

The most important implementation detail is that a strong algorithm can still be deployed badly. Low iteration counts, weak parameter choices, or preserving old hashes indefinitely can erase much of the intended benefit even when the algorithm itself is modern.

What to choose when compatibility, migration, or legacy constraints matter

In a new design, choose a memory hard function first, then tune the parameters for your production latency budget and expected authentication volume. If you are maintaining a legacy system, PBKDF2 is often the compatibility fallback because it is broadly supported and straightforward to migrate toward a stronger scheme later. NHIMG’s Password Hashing and Key Derivation Guide covers the practical decision points, including parameter tuning, peppers, and migration from older hashes.

When the question is operational rather than theoretical, the right answer is usually hybrid: keep existing verified hashes until users authenticate, then upgrade them in place to the stronger algorithm you have standardised on. That approach reduces forced resets while still shrinking the attack window over time.

Hashing Guide is useful when you need the wider context for salts, hash properties, and why password hashing differs from general purpose hashing. For teams setting policy, Password Security and Password Manager Guide helps connect password storage choices to the broader credential lifecycle and password policy decisions.

Risk and Threat Considerations

Password hashes are usually attacked offline after a database leak, so the main risk is not login abuse but bulk cracking at scale. The weaker the work factor and the less memory the function requires, the more attractive the stored hashes become to attackers with specialised hardware.

Failure mechanism: Adaptive functions that rely mainly on repeated computation can be spread efficiently across modern GPU clusters, while weak tuning or legacy parameter choices lowers the cost per guess and shortens the time to recover real passwords.

Impact: Cracked passwords can be reused for account takeover, lateral access, and credential stuffing across other services, especially where users reuse passwords or where the compromised system stores privileged accounts.

Standards & Framework Alignment

This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.

NIST SP 800-53 Rev 5, NIST SP 800-57 and CIS Controls v8 set the governance and control requirements practitioners need to meet.

FrameworkControl / ReferenceRelevance
NIST SP 800-53 Rev 5IA-5 — Authenticator ManagementPassword storage depends on managing verifier strength and lifecycle.
Recommendation — Set and enforce strong password hashing parameters and rotation rules for stored authenticators.
NIST SP 800-57Key ManagementPeppering and related secret handling tie password storage to cryptographic secret lifecycle.
Recommendation — Protect peppers and related secrets with strong lifecycle controls and restricted access.
CIS Controls v8CIS-5 — Account ManagementStored password strength affects account compromise risk and recovery.
Recommendation — Use strong password storage to reduce account takeover risk from database theft.

Practitioner Guidance

What to verify: Confirm that the live hash format, work factor, and parameter set are actually the ones you expect in production. Many teams believe they have “migrated” when only new accounts are using the stronger function and old hashes are still widely present.

Decision rule: If the system can support a memory hard function without unacceptable authentication latency, prefer it for new password storage. If compatibility forces PBKDF2, treat that as a temporary constraint and set a clear migration trigger instead of leaving it as the permanent default.

Practitioner takeaway: The real security difference is not just the algorithm family, but whether your deployment makes offline cracking expensive enough that modern hardware no longer gives attackers a large advantage.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Reviewed and updated by the NHIMG editorial team on October 31, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org