As per my understanding both JWT and Basic Auth used to store login credentials on client side and avoid sessions for better scalability. I understand with Basic Auth login credentials will be sent along with each request which is a security risk incase of http but with https these credentials will be encrypted which prevents from eavesdropping.Even with JWT token we are sending user credentials with each request. So what are the advantages of using JWT over basic auth with https?

  • 2
    "Even with JWT token we are sending user credentials with each request." - usually not. Can you please add some reference which claims that JWT have to be used this way or are commonly used this way? Given that your question is based on this claim it would make the whole question obsolete if the claim is not true, right? Apr 11, 2021 at 17:25
  • 1
    For starters, JWT provides the integrity check for the claim made by user where the basic authentication can be sent by anyone once the values are known which has potential for replay and other similar attacks.
    – nethero
    Apr 11, 2021 at 21:02

2 Answers 2


First, as Steffen's comment points out, you don't send credentials (certainly not complete ones, suitable for authentication) in JWTs. I mean, you could, but that both completely misses the point of a JWT and is totally useless. In my years of reviewing many, many web apps and services I have never seen this pattern. I'm really curious where you got the notion that this is a thing people do.

JWTs contain three parts. There's metadata (information about the JSON Web Token), like who issued it and when it is / isn't valid. There's claims (information about the token bearer), which in the most basic form is one field containing an email address or similar identifying string (or sometimes number). Finally there's the signature, which can only be created with knowledge of a secret (either a symmetric HMAC key or an asymmetric digital signature private key) that the user doesn't have, and which prevents the user (or anybody else) from forging or modifying the JWT and having the server still accept it.

Note the total lack in there of anything a user could use to prove their identity (a "credential"). No passwords (as there are in HTTP Basic Auth). No ephemeral user access codes, such as a random string from an email or app or SMS. No sign of user possession of a cryptographic key (such as is used in TLS mutual auth; the only cryptographic key used in a JWT is exclusive to the server). Yes, the claims may contain a user identifier such as a username, but so does the question you posted, the "From" field of any email you send or receive, and the sender info of every SMS. You can't use those (alone) to log in to anything. They aren't credentials, just identifiers. Access to the account behind the email address or username, or the phone behind the number, can be used as a credential, but the identifier itself is not.

So, what are the advantages of JWTs over HTTP Basic Auth? In no particular order:

  • JWTs are generally specific to one service. If I take a JWT that tells StackExchange that I'm cbhacking and try to use it on GitHub, there is no chance GitHub will consider this valid. Passwords (sadly) get reused all over.
  • JWTs are time-limited, and expire. If you dig through a dump of some poorly-sanitized logs from months ago and find a JWT I used, it will almost certainly (if the implementer had any sense) long since have expired and thus be basically worthless. If you instead find a basic auth string, odds are pretty good that the password it contains is still valid.
  • JWTs can be issued by one service and verified by another. This is one of the standard things they're used for, in fact. If you log into a site A via "Sign in with Facebook" or some other SSO (Single Sign-On) provider B, the site A that you're logging into will never see your password; that either goes straight to the SSO provider B, or did at some time in the past (and your session with B is still active). Then B can return a JWT saying "The bearer of this is Arjun, user ID 255431" and signed in a way that A can verify but not create (this is the magic of asymmetric cryptography). Now site A knows who you are - so long as you send the JWT to them - without ever seeing any of your credentials.
  • JWTs can grant scoped permissions. Suppose I have a calendar online, and since it's my calendar, I can see all the details of events, add and remove events, and so on. Now suppose some third-party app for scheduling calls with clients wants to access my calendar. I want it to know what times I'm available, but I don't want it to know what I'm doing during any particular slot. Maybe I'd let it add and remove those client meetings directly, but it shouldn't be able to remove meetings (or dentist appointments, or vacations, or friends' birthdays) that I added for my own use (nor should it even be able to tell what those are). With basic auth, I'd have to set up multiple unique passwords - one for each service (client meetings, friend birthdays, etc.) that I don't want to give full access to - or else share the main password and then any app that wanted to could take total control. With a JWT, each app can request only the permissions that it needs, and when I approve them it gets a JWT with "scope" claims that limit what the token can be used for.
  • JWTs can be both secure and fast to verify. Password storage, necessary for basic auth, can't. Symmetric signatures (HMACs) on JWTs are super-fast to verify, usually at most taking a few rounds of hashing. Asymmetric signatures are significantly more expensive to verify, but you can still do them very quickly especially if you use efficient algorithms, without compromising security. Verifying passwords, however, needs to be very slow indeed, because they're too easy to brute-force from a leaked database if you use a fast hash or similar. Thus password hashing is deliberately designed to be computationally expensive - tens of thousands of rounds of hashing is one simple way to do it - and is usually tuned to take somewhere between about 50ms and 250ms per user, on security-conscious sites. As computers get faster, you have to make the password hash even more expensive, too, which isn't a significant concern with JWT signatures. This means that, with basic auth, you have to either add a ton of CPU time spent doing extremely expensive password hashing, or you have to run a very high risk that, if your password database is leaked, people will be able to brute-force the hashes. With JWTs that's not nearly as big a concern (with HMAC-signed JWTs, it's basically not a concern at all).
  • Relatedly, JWTs can be verified without a database lookup at all. Passwords, including ones sent in basic auth, can't. You can cache them, but that takes server-side resources and breaks down when you have a cluster of servers behind a load balancer. Thus, JWTs are much better for scaling to large numbers of users, where DB hits are relatively expensive and need to be minimized.
  • Basic auth isn't really compatible with multi-factor authentication (sometimes called two-step authentication or similar). MFA is usually implemented with ephemeral values - one-time codes generated in an app or sent via SMS, challenge nonces sent to hardware tokens or push notifications, etc. - and even if there was a practical way to include the resulting token in the basic auth string (there isn't, for browsers), it would become invalid almost immediately. That's awkward if you want a session to last more than a few minutes before the user has to log in again! With a JWT (or any other session token), you can verify the user's credentials (including MFA) once, and then give them a token good for as long as you want (with JWTs in particular, it's often two tokens: the JWT itself which has a short lifetime, and a refresh token that lasts longer but takes a DB lookup to verify).
  • A bit of a technical curiosity rather than a deliberate feature of JWT design, but: HTTP Basic Auth is automatically sent by the browser for any request to the relevant domain... even if the request comes from a different domain. This means that there's a risk of CSRF (Cross-Site Request Forgery), where if site X thinks you might be signed into site Y, it can make your browser send requests to Y and Y will think you, the user, did it. This could be used for anything from posting messages on a forum to transferring money between accounts, and it would happen without your awareness. JWTs, by comparison, can be put in lots of different places. Cookies are common, and used to be automatically sent even on requests from other domains (just like Basic Auth), but these days there's the SameSite cookie flag that can be used to control this behavior. JWTs are also often sent in other HTTP request headers (often Authorization - the same one used for basic auth - but sometimes custom ones), and in those cases the requesting client must explicitly specify to send the JWT. Which means the requesting client has to know the full contents of the JWT, because your browser won't just automatically attach it to every relevant request.
  • Logging out a user who uses a JWT is simple, just delete it from their client (this doesn't invalidate the JWT on the server - that's hard to do, and usually you just rely on the JWT having a short lifetime and expiring soon - but it works fine for Alice getting up from the computer before Bob sits down to use it). It turns out that there's no standardized way to make a browser forget basic auth credentials. At least one browser (Internet Explorer) has a special API for this, but generally speaking, if you want to log out a client that uses basic auth, the only option is to tell the browser that their authorization is invalid (which generally makes the browser stop sending it). That's... messy. Obviously for non-browser clients this is not an issue.
  • Relatedly, logging in with basic auth isn't really under the control of the client application (implemented in HTML/CSS/JS). The web browser chooses what to show (usually a pretty ugly modal credentials prompt). Other forms of authentication / authorization token allow the web application to create their own login pages, which might include features like a "forgot password?" link or SSO login options, while a basic auth site just presents the user with a blank page (the content won't be loaded - or even requested - until the user signs in!) and an ugly pop-up modal box.
  • This is a pretty good explanation of the topic. I always got confused about what's even the point of having JWT's if they're being stored on the client but I guess this kinda clears it up along with all the various benefits. Aug 7, 2022 at 3:27
  • Thank you so much for such a great answer
    – dbnex14
    Jan 18, 2023 at 19:53

There are a couple different approaches one can take to authentication. In your question, you mention Basic authentication. If this is a username and a user-chosen password, then the downside to this is that users pick bad passwords, and ultimately in many cases, those are easy to guess and frequently reused. Thus, any solution which avoids user-generated passwords is generally an improvement in security. In addition, user passwords do not change frequently, so if they are exposed, the compromise can last a long time.

If you're using Basic authentication with a username or identifier and a randomly generated token, then the tradeoffs are different with a JWT, and you can make a decision about the best option based on your needs.

A JWT (and similar signed tokens) provide a way to embed information, which in the JWT context are called claims, that are authenticated, usually with a digital signature or HMAC key, and may also be encrypted. This means that one service can perform authentication and issue tokens which can be accepted by other services without the need to have the two services communicate; all the latter service needs to do is have a way to verify the signing credentials. However, this also has the downside that if you do this and the credentials are exposed, there's generally no way to revoke them and they remain valid until they expire.

The benefit to using a token which is just randomly generated with a CSPRNG or which only embeds a simple signed identifier is that at any point, if the credential is revoked, it will be noticed nearly immediately. These credentials also tend to be smaller than JWTs and similar tokens, since the access is verified via the authentication system and not embedded in the token. However, they require all services to be in some sort of contact with an authentication service to verify credentials, which may not be suitable for your system's architecture.

Which one to use depends on the architecture of your system and your needs. Like all things in security, there are tradeoffs, in this case between security and scalability. In both cases, you should try to make the life of your tokens as short as possible. If you do need to issue long-term tokens to users, try to use a different, shorter length token in places where you can.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .