Assume the following scenario: we want to implement an open-source password manager that uses a central service that enables the different clients (on different devices) to synchronize their local databases. It doesn't matter if this service is run by a company or on a server of the user (compare to owncloud usage scenarios). To make our application more "secure", we want to use an Intel SGX enclave for the central service (please ignore current attack research on SGX enclaves).
Then, the typical workflow would be:
local client attests central enclave
user registers / logs in
(local and remote database are synced)
user stores / retrieves passwords
Now my question: Does every user of our password manager need to register with the Intel Attestation Service (IAS)? If yes, wouldn't this imply that, since private key sharing is really bad, every single device needs to be registered?
According to my investigations, the answer is, at least for the development and testing phase, yes. I could not find any public information for production scenarios. All I know is that a business registration changes the behavior of the enclave (it can be run in production mode) which is not of any matter here. I have two thoughts on this:
If I am correct, isn't this another show stopper for SGX? Using SGX without the attestation feature seems to be useless.
How do services such as https://www.fortanix.com/ circumvent or solve the problem? Their documentation does not give a hint for needed interaction with Intel.
The above described scenario is only an example; it can be improved and we do not plan to implement it. But it was a lot easier to describe a scenario, that can be easy imagined and seems to be a realistic use case for SGX, than describing our current project plans.
P.S.: This question is kind of consecutive to Intel SGX developer licensing and open-source software
One does not need an Intel-registered certificate to create a quote but one does need to communicate with the IAS (Intel Attestation Service) to verify a quote, which requires an Intel-registered certificate. So every node checking if a remote attestation is valid would require such a certificate in a naive approach.
One could of course leverage SGX to provide a proxy which would be structured somewhat like that:
Generate two certificates and their corresponding private key, I'll name one of them the IAS-conn-cert and the other one the Proxy-cert.
Register the IAS-conn-cert of them to the IAS.
Of course, you need to have to trust that these certificates were indeed generated in an enclave. To do so, you could remotely attest to another service provider you trust.
Now pin (by for example hard-coding) the Proxy-cert in your client application. When it needs to verify a quote, it connects to the enclave using that pinned proxy-cert thus knowing it connects to the enclave. The enclave will then connect to the IAS and relay everything it receives from the client to the IAS and vise-versa. The client can now communicate with the IAS without having to own an IAS-registered certificate but can still be assured that there is no tampering in the proxy, given that it trusts that the proxy-certificate was indeed generated in a non-malicious enclave.
I'd like to use Windows.Security.Credentials.PasswordVault in my desktop app (WPF-based) to securely store a user's password. I managed to access this Windows 10 API using this MSDN article.
I did some experiments and it appears that any data written to PasswordVault from one desktop app (not a native UWP app) can be read from any other desktop app. Even packaging my desktop app with Desktop Bridge technology and thus having a Package Identity does not fix this vulnerability.
Any ideas how to fix that and be able storing the app's data secure from other apps?
UPDATE: It appeared that PasswordVault adds no extra security over DPAPI. The case is closed with a negative result.
(this is from what I can understand of your post)
There is no real way of preventing data access between desktop apps when using these kind of API's http://www.hanselman.com/blog/SavingAndRetrievingBrowserAndOtherPasswords.aspx tells more about it. You'd probably just want to decrypt your information.
memory access restriction is difficult, code executed by the user is always retrievable by the user so it would be difficult to restrict this.
have you considered using the Windows Data Protection API :
https://msdn.microsoft.com/en-us/library/ms995355.aspx
grabbed straight from the source
DPAPI is an easy-to-use service that will benefit developers who must provide protection for sensitive application data, such as passwords and private keys
WDPAPI uses keys generated by the operating system and Triple DES to encrypt/decrypt your data. Which means your application doesn't have to generate these keys, which is always nice.
You could also use the Rfc2898DeriveBytes class, this uses a pseudo-random number generator to decrypt your password. It's safer than most decrypters since there is no practical way to go back from the result back to the password. This is only really useful for verifying the input password and not retrieving it back again. I have never actually used this myself so I would not be able to help you.
https://msdn.microsoft.com/en-us/library/system.security.cryptography.rfc2898derivebytes(v=vs.110).aspx
see also this post which gives a way better explanation than I can.
How to securely save username/password (local)?
If I misunderstood the question in some way, tell me, I will try to update the answer.
NOTE that modern/metro apps do not have this problem, although they still are accessible in other ways.
The hard truth is that storing a password in a desktop application, 100% securely is simply not possible. However, you can get close to 100%.
Regarding your original approach, PasswordVault uses the Credential Locker service which is built into windows to securely store data. Credential Locker is bound to the user's profile. Therefore, storing your data via PasswordVault is essentially equivalent to the master password approach to protecting data, which I talk about in detail further down. Only difference is that the master password in that case is the user's credentials. This allows applications running during the user's session to access the data.
Note: To be clear, I'm strictly talking about storing it in a way that allows you access to the plain text. That is to say, storing it in an encrypted database of any sort, or encrypting it yourself and storing the ciphertext somewhere. This kind of functionality is necessary in programs like password managers, but not in programs that just require some sort of authentication. If this is not a necessity then I strongly recommend hashing the password, ideally per the instructions laid out in this answer by zaph. (Some more information in this excellent post by Thomas Pornin).
If it is a necessity, things get a bit more complicated: If you want to prevent other programs (or users I suppose) from being able to view the plaintext password, then your only real option is to encrypt it. Storing the ciphertext within PasswordVault is optional since, if you use good encryption, your only weak point is someone discovering your key. Therefore the ciphertext itself can be stored anywhere. That brings us to the key itself.
Depending on how many passwords you're actually trying to store for each program instance, you might not have to worry about generating and securely storing a key at all. If you want to store multiple passwords, then you can simply ask the user to input one master password, perform some salting and hashing on that, and use the result as the encryption key for all other passwords. When it is time for decryption, then ask the user to input it again. If you are storing multiple passwords then I strongly urge you to go with this approach. It is the most secure approach possible. For the rest of my post however, I will roll with the assumption that this is not a viable option.
First off I urge you not to have the same key for every installation. Create a new one for every instance of your program, based on securely generated random data. Resist the temptation to "avoid having to store the key" by having it be generated on the fly every time it is needed, based on information about the system. That is just as secure as hardcoding string superSecretKey = "12345"; into your program. It won't take attackers long to figure out the process.
Now, storing it is the real tricky part. A general rule of infosec is the following:
Nothing is secure once you have physical access
So, ideally, nobody would. Storing the encryption keys on a properly secured remote server minimizes the chances of it being recovered by attackers. Entire books have been written regarding server-side security, so I will not discuss this here.
Another good option is to use an HSM (Hardware Security Module). These nifty little devices are built for the job. Accessing the keys stored in an HSM is pretty much impossible. However, this option is only viable if you know for sure that every user's computer has one of these, such as in an enterprise environment.
.Net provides a solution of sorts, via the configuration system. You can store your key in an encrypted section of your app.config. This is often used for protecting connection strings. There are plenty of resources out there on how to do this. I recommend this fantastic blog post, which will tell you most of what you need to know.
The reason I said earlier not to go with simply generating the key on the fly is because, like storing it as a variable in your code, you rely exclusively on obfuscation to keep it secure. The thing about this approach is that it usually doesn't. However, sometimes you have no other option. Enter White Box cryptography.
White box cryptography is essentially obfuscation taken to the extreme. It is meant to be effective even in a white-box scenario, where the attacker both has access to and can modify the bytecode. It is the epitome of security through obscurity. As opposed to mere constant hiding (infosec speak for the string superSecretKey approach) or generating the key when it is needed, white box cryptography essentially relies on generating the cipher itself on the fly.
Entire papers have been written on it, It is difficult to pull off writing a proper implementation, and your mileage may vary. You should only consider this if you really really really want to do this as securely as possible.
Obfuscation however is still obfuscation. All it can really do is slow the attackers down. The final solution I have to offer might seem backwards, but it works: Do not hide the encryption key digitally. Hide it physically. Have the user insert a usb drive when it is time for encryption, (securely) generate a random key, then write it to the usb drive. Then, whenever it is time for decryption, the user only has to put the drive back in, and your program reads the key off that.
This is a bit similar to the master password approach, in that it leaves it up to the user to keep the key safe. However, it has some notable advantages. For instance, this approach allows for a massive encryption key. A key that can fit in a mere 1 megabyte file can take literally billions of years to break via a brute force attack. Plus, if the key ever gets discovered, the user has only themselves to blame.
In summary, see if you can avoid having to store an encryption key. If you can't, avoid storing it locally at all costs. Otherwise, your only option is to make it as hard for hackers to figure it out as possible. No matter how you choose to do that, make sure that every key is different, so even if attackers do find one, the other users' keys are safe.
Only alternative is to encrypt password with your own private key stored somewhere in your code. (Someone can easily disassemble your code and get the key) and then store encrypted password inside PasswordVault, however the only security you have is any app will not have access to password.
This is dual security, in case of compromised machines, attacker can get access to PasswordVault but not your password as they will need one more private key to decrypt the password and that will be hidden somewhere in your code.
To make it more secure, if you leave your private key on your server and expose an API to encrypt and decrypt password before storing in Vault, will make it most secure. I think this is the reason people have moved on to OAuth (storing OAuth token in PasswordVault) etc rather then storing password in vault.
Ideally, I would recommend not storing password, instead get some token from server and save it and use that token for authentication. And store that token in PasswordVault.
It is always possible to push the security, with miscellaneous encryption and storage strategies. Making something harder is only making the data retrieval longer, never impossible. Hence you need to consider the most appropriate level of protection considering execution cost x time (human and machine) and development cost x time aspects.
If I consider strictly your request, I would simply add a layer (class, interface) to cipher your passwords. Best with asymmetrical encryption (and not RSA). Supposing the other softs are not accessing your program data (program, files OR process), this is sufficient. You can use SSH.NET (https://github.com/sshnet/SSH.NET) to achieve this quickly.
If you would like to push the security and give a certain level of protection against binary reverse-engineering (including the private key retrieval), I recommend a small (process limited) encrypted VM (like Docker, https://blogs.msdn.microsoft.com/mvpawardprogram/2015/12/15/getting-started-with-net-and-docker/) based solution such as Denuvo (https://www.denuvo.com/). The encryption is unique per customer and machine based. You'll have to encapsulated you c# program into a c/c++ program (which acts like a container) that will do all the in-memory ciphering-deciphering.
You can implement your own strategy, depending on the kind of investment and warranty you require.
In case your program is a backend program, you can pick the best strategy (the only I really recommend) of all which is to store the private key at the client side, public key at backend side and have local deciphering, all transmitted password would be hence encrypted. I would like to remark that password and keys are actually different strategies to achieve the same goal: checking if the program talks to the right person without knowing the person's identity; I mean this: instead of storing passwords, better store directly public keys.
Revisiting this rather helpful issue and adding a bit of additional information which might be helpful.
My task was to extend a Win32 application that uses passwords to authenticate with an online service with a "save password" functionality. The idea was to protect the password using Windows Hello (UserConsentVerifier). I was under the impression that Windows surely has something comparable to the macOS keychain.
If you use the Windows Credential Manager APIs (CredReadA, CredWriteA), another application can simply enumerate the credentials and if it knows what to look for (the target name), it will be able to read the credential.
I also explored using DPAPI where you are in charge of storing the encrypted blob yourself, typically in a file. Again, there seems to be no way (except obfuscation) to prevent another application from finding and reading that file. Supplying additional entropy to CryptProtectData and CryptUnprotectData again poses the question of where to store the entropy (typically I assume it would be hard-coded and perhaps obfuscated in the application: this is security by obscurity).
As it turns out, neither DPAPI (CryptProtectData, CryptUnprotectData) nor Windows Credential Manager APIs (CredRead, CredWrite) can prevent another application running under the same user from reading a secret.
What I was actually looking for was something like the macOS keychain, which allows applications to store secrets, define ACLs on those secrets, enforce biometric authentication on accessing the secret, and critically, prevents other applications from reading the secrets.
As it turns out, Windows has a PasswordVault which claims to isolate apps from each other, but its only available to UWP apps:
Represents a Credential Locker of credentials. The contents of the locker are specific to the app or service. Apps and services don't have access to credentials associated with other apps or services.
Is there a way for a Win32 Desktop application to access this functionality? I realize that if a user can be brought to install and run a random app, that app could probably mimic the original application and just prompt the user to enter the secret, but still, it's a little disappointing that there is no app-level separation by default.
Setup
I have a SQLite database which has confidential user information.
This database may be replicated on other machines
I trust the user, but not other applications
The user has occasional access to a global server
Security Goals
Any program other than the authorized one (mine) cannot access the SQLite database.
Breaking the security on one machine will NOT break the security on other machines
The system must be updatable (meaning that if some algorithm such as a specific key generation algorithm is shown to be flawed, it can be changed)
Proposed Design
Use an encrypted SQLite database storing the key within OS secure storage.
Problems
Any windows hack will allow the person to access the key for all machines which violates goal #2
Notes
Similar to this method, if I store the key in the executable, breaking the security will comprimise all systems.
Also, I have referenced windows secure storage. While, I will go to an os specific solution if I have to, I would prefer a non-os specific solution
Any idea on how to meet the design goals?
I think you will need to use TPM hardware e.g. via TBS or something similar, to actually make a secure version of this. My understanding is, TPM lets the application check that it is not being debugged or traced at a software level, and the operating system should prevent any other application pretending to the TPM module that it is your application. I may be wrong though.
You can use some kind of security-through-obscurity kludge, but it will be crackable with a debugger unless you use TPM.
Suppose I need a login mechanism for a program in a Local Area Network in a company, my guess is to store a file with username/password pairs on the local server, but would the Java program be able to read/write information to the file from a local PC? It's my first time dealing with such a task so I am a bit confused about this. Also I want to store only the passwords for the program, not the PC user.
Hmm, you should do it differently imho.
Write a service to authenticate against. The service is the only application allowed to read the password-file and runs on the server. The clients authenticate against that service. Once the user is authenticated, pass him an identification token that is tied to his machine and can expire after a period. Also, the machine needs to transmit some sort of digital signature to verify its integrity in an asynchronous manner. If you do this, you can verify that only authenticated users, who really are who they claim to be can access services which require the authentication token, including the authentication service itself.
BUT: I strongly suggest you get something that has already been built for such tasks. There're things like Kerberos which have been built for such tasks. I am not a sysadmin, you might ask again at serverfault or so.
Additionally, I'd like to state that MD5 is not the toughest hash anymore. AFAIK blowfish is the way to go today, I might be wrong, though. It's tougher than MD5 anyway, which is prone to collision-attacks already.
Is using a USB key to secure an application the best option?
If it isn't, what is the best way to secure an application in the form of requiring a valid user before the application can be used?
The reason I ask this question is that a client recently asked me to make an application require a specific USB device be inserted into the system before the app can be used or run. Basically, the application is a medical software and the client sees this method of security more trusted, as the app contains sensitive data of his paients. In that case of loss, I can tell the user how to generate another security key using the app in command line mode.
Not necessarily an answer to the question, but a point to be considered in response to the question and comments so far...
A USB key by itself is not going to be more secure than password-based authentication. It's still one-factor, it can be lost/stolen/etc. What the client probably actually wants, and either doesn't know it or hasn't properly articulated it, is multi-factor authentication. Consider these:
Something you know (a password, answer to a question, etc.)
Something you have (a USB key, time-limited key generating fob, etc.)
Something you are (finger print, retina scan, etc.)
Most systems use only the first one. For added security, you add the second one. For Mission: Impossible style high-clearance security, throw in the third one. The idea is that any one factor can be forged, but adding new factors adds new dimensions of security which exponentially make it more difficult (rather than linearly more difficult when just replacing one factor with a "better" factor).
Your application could require a certificate, signed by you, to be present on a regular USB memory stick. The certificate could be locked with a password, that the user would have to enter. Certificates on stolen or lost thumbdrives could be revoked.
For a less homebrew solution, I'm sure there are commercial smart card based services you could use.