Best practice to detect iPhone app only access for web services? - iphone

I am developing an iPhone app together with web services. The iPhone app will use GET or POST to retrieve data from the web services such as http://www.myserver.com/api/top10songs.json to get data for top ten songs for example.
There is no user account and password for the iPhone app. What is the best practice to ensure that only my iPhone app have access to the web API http://www.myserver.com/api/top10songs.json? iPhone SDK's UIDevice uniqueueIdentifier is not sufficient as anyone can fake the device id as parameter making the API call using wget, curl or web browsers.
The web services API will not be published. The data of the web services is not secret and private, I just want to prevent abuse as there are also API to write some data to the server such as usage log.

What you can do is get a secret key that only you know, Include that in an md5 hashed signature, typically you can structure signatures as a s tring of your parameters a nd values and the secret appended at the end, then take the md5 hash of that...Do this both in your client and service side and match the signature string, only if the signatures match do you get granted access...Since t he secret is only present i n the signature it w ill be hard to reverse engineer and crack..

Here's an expansion on Daniel's suggestion.
Have some shared secret that the server and client know. Say some long random string.
Then, when the client connects, have the client generate another random string, append that to the end of the shared string, then calculate the MD5 hash.
Send both the randomly generated string and the hash as parameters in the request. The server knows the secret string, so it can generate a hash of its own and make sure it matches the one it received from the client.
It's not completely secure, as someone could decompile your app to determine the secret string, but it's probably the best you'll get without a lot of extra work.

Use some form of digital signatures in your request. While it's rather hard to make this completely tamper proof (as is anything with regard to security). It's not that hard to get it 'good enough' to prevent most abuse.
Of course this highly depends on the sensitivity of the data, if your data transactions involve million dollar transactions, you'll want it a lot more secure than some simple usage statistic logging (if it's hard enough to tamper and it will gain little to no gain to the attacker except piss you of, it's safe to assume people won't bother...)

I asked an Apple security engineer about this at WWDC and he said that there is no unassailable way to accomplish this. The best you can do is to make it not worth the effort involved.
I also asked him about possibly using push notifications as a means of doing this and he thought it was a very good idea. The basic idea is that the first access would trigger a push notification in your server that would be sent to the user's iPhone. Since your application is open, it would call into the application:didReceiveRemoteNotification: method and deliver a payload of your own choosing. If you make that payload a nonce, then your application can send the nonce on the next request and you've completed the circle.
You can store the UDID after that and discard any requests bearing unverified UDIDs. As far as brute-force guessing of necessary parameters, you should be implementing a rate-limiting algorithm no matter what.

A very cheap way to do this could be getting the iPhone software to send extra data with the query, such as a long password string so that someone can't access the feed.
Someone could reverse engineer what you have done or listen to data sent over the network to discover the password and if bandwidth limitations are the reason for doing this, then a simple password should be good enough.
Of course this method has it's problems and certificate based authentication will actually be secure, although it will be harder to code.

The most secure solution is probably a digital signature on the request. You can keep a secret key inside the iPhone app, and use it to sign the requests, which you can then verify on the server side. This avoids sending the key/password to the server, which would allow someone to capture it with a network sniffer.
A simple solution might be just to use HTTPS - keeping the contents of your messages secure despite the presence of potential eavesdroppers is the whole point of HTTPS. I'm not sure if you can do self-signed certificates with the standard NSURLConnection stuff, but if you have a server-side certificate, you're at least protected from eavesdropping. And it's a lot less code for you to write (actually, none).
I suppose if you use HTTPS as your only security, then you're potentially open to someone guessing the URL. If that's a concern, adding just about any kind of parameter validation to the web service will take care of that.

The problem with most if not all solutions here is that they are rather prone to breaking once you add proxies in the mix. If a proxy connects to your webservice, is that OK? After all, it is probably doing so on behalf of an iPhone somewhere - perhaps in China? And if it's OK for a proxy to impersonate an iPhone, then how do you determine which impersonations are OK?

Have some kind of key that changes every 5 minutes based on an algorithm which uses the current time (GMT). Always allow the last two keys in. This isn't perfect, of course, but it keeps the target moving, and you can combine it with other strategies and tactics.
I assume you just want to dissuade use of your service. Obviously you haven't set up your app to be secure.

Related

iPhone App with a Server Backend - How to ensure all access is from the iPhone app only?

I don't mind so much about pirating etcetera, but I want to ensure that the backend (Rails based) isn't open to automated services that could DOS it etc. Therefore I'd like to simply ensure that all access to the backend (which will be a few REST queries to GET and PUT data) will be via a valid iPhone application, and not some script running on a machine.
I want to avoid the use of accounts so that the user experience is seamless.
My first intention is to hash the UDID and a secret together, and provide that (and the UDID) over a HTTPS connection to the server. This will either allow an authenticated session to be created or return an error.
If eavesdropped, then an attacker could take the hash and replay it, leaving this scheme open to replay attacks. However shouldn't the HTTPS connection protect me against eavesdropping?
Thanks!
Like bpapa says, it can be spoofed, but then, like you say, you aren't worried about that so much as anybody coming along and just sending a thousand requests to your server in a row, and your server having to process each one.
Your idea of the hash is a good start. From there, you could also append the current timestamp to the pre-hashed value, and send that along as well. If the given timestamp is more than 1 day different from the server's current time, disallow access. This stops replay attacks for more than a day later anyway.
Another option would be to use a nonce. Anybody can request a nonce from your server, but then the device has to append that to the pre-hash data before sending the hash to the server. Generated nonces would have to be stored, or, could simply be the server's current timestamp. The device then has to append the server's timestamp instead of its own timestamp to the pre-hashed data, allowing for a much shorter period than a full day for a replay attack to occur.
Use SSL with client certificate. Have a private key in your client and issue a certificate for it, and your web server can require this client cert to be present in order for the sessions to proceed.
I can't give code details for Rails, but architecture-wise it's the most secure thing to do, even though might be a bit overkill. SSL with certificates is a standard industry solution and libraries exist for both the iPhone/client end and server end, so you don't have to invent anything or implement much, just get them to work nicely together.
You could also consider HMAC, like HMAC-SHA1, which is basically a standardization of the hashes stuff that other people here talk about. If you added nonces to it, you'd also be safe against replay attack. For an idea about how to implement HMAC-SHA1 with nonces, you could look at OAuth protocol (not the whole flow, but just how they tie nonce and other parameters together into an authenticated request).
There is no way to ensure it, since it can be spoofed.
If you really want to go this route (honestly, unless you're doing something really super mission critical here you are probably wasting your time), you could pass along the iPhone device token. Or maybe hash it and then pass it along. Of course, you have no way to validate it on the Server Side or anything, but if a bad guy really wants to take you down, here is roadblock #1 that he will have to deal with first.

Is this plan for preventing iPhone app client spoofing sound?

I'm designing an iPhone app that communicates with a server over HTTP.
I only want the app, not arbitrary HTTP clients, to be able to POST to certain URL's on the server. So I'll set up the server to only validate POSTs that include a secret token, and set up the app to include that secret token. All requests that include this token will be sent only over an HTTPS connection, so that it cannot be sniffed.
Do you see any flaws with this reasoning? For example, would it be possible to read the token out of the compiled app using "strings", a hex editor, etc? I wouldn't be storing this token in a .plist or other plain-text format, of course.
Suggestions for an alternate design are welcome.
In general, assuming that a determined attacker can't discover a key that is embedded in application on a device under his physical control (and, probably, that he owns anyway) is unwarranted. Look at all of the broken DRM schemes that relied on this assumption.
What really matters is who's trying to get the key, and what their incentive is. Sell a product aimed at a demographic that isn't eager to steal. Price your product so that it's cheaper to buy it than it is to discover the key. Provide good service to your customers. These are all marketing and legal issues, rather than technological.
If you do embed a key, use a method that requires each client to discover the key themselves, like requiring a different key for each client. You don't want a situation where one attacker can discover the key and publish it, granting everyone access.
The iPhone does provide the "KeyChain" API, which can help the application hide secrets from the device owner, for better or worse. But, anything is breakable.
The way I understand it, yes, the key could be retrieved from the app one way or another. It's almost impossible to hide something in the Objective-C runtime due to the very nature of it. To the best of my knowledge, only Omni have managed it with their serial numbers, apparently by keeping the critical code in C (Cocoa Insecurity).
It might be a lot of work (I've no idea how complex it is to implement), but you might want to consider using the push notifications to send an authentication key with a validity of one hour to the program every hour. This would largely offload the problem of verifying that it's your app to Apple.
I suggest to add some checksum (md5/sha1) based on the sent data and a secret key that your app and the server knows.
Applications can be disassembled so that they could find your key.
More information is needed to determine whether the approach is sound. It may be sound for one asset being protected and unsound for another, all based on the value of the asset and the cost if the asset is revealed.
Several earlier posters have alluded to the fact that anything on the device can be revealed by a determined attacker. So, the best you can do is determine valuable the asset is and put enough hurdles in the way of the attacker that the cost of the attack exceeds the value of the asset.
One could add to your scheme client-side certificates for the SSL. One could bury that cert and the key for the token deep in some obfuscated code. One could probably craft a scheme using public/private key cryptography to further obscure the token. One could implement a challenge/response protocol that has a time boxed response time wherein the server challenges the app and the app has X milliseconds to respond before it's disconnected.
The number and complexity of the hurdles all depend on the value of the asset.
Jack
You should look into the Entrust Technologies (www.entrust.com) product line for two-factor authentication tied to all sorts of specifics (e.g., device, IMEI, application serial number, user ID, etc.)

How to ensure access to my web service from my code only?

I am writing a very simple web service for my iPhone app. Let's say this is a http page that returns a random number at http://mysite/getRand. How do I ensure that this page can only be accessed from my iPhone app and not from other clients? I've thought of doing some simple password mechanism but that can easily be sniffed by capturing what my app sends out.
The reason for this is to lower the load of my server by only allowing legitimate requests.
You can't really do this. Your application can be disassembled and whatever secret is in the binary can be replicated in a malicious application.
Another attack you should be aware of is people settings the hosts file to a location they control and then installing a root certificate that allows them to provide a signature for that domain. Your application would do the post with the secret, and they'd just be able to read out the secret. They could extract the password from any complicated encryption system within the binary in this way.
Most of the ideas in this thread are vulnerable to this attack.
That said, the likelihood of somebody caring enough to disassemble your application is probably fairly remote.
I'd just keep it simple. Have a password that's hardcoded in to your application. To prevent someone just looking at the resources and trying every string, make it the XOR of two strings or the result of an AES decrypt of a particular fixed string.
Obviously, you should do the request over SSL otherwise an attacker can just sniff the traffic.
Yes, a determined attacker will circumvent the scheme but like any DRM scheme, that's always been the case. The trick is to make it too much effort to be worth it.
To follow up on Simon's idea, you could very easily have a key string in your application, then send the device ID, and then the DeviceID XOR'ed (or some other simple algorithm for string encryption) with your key string.
Since you know the key value to use, it's trivial for you to "decrypt" this string on the sever side and verify that the values match.
This way, the password is different for each user's device, and the "key" string is never sent over the wires of the great unwashed internets. :-)
Yes, this would by no means be impossible to figure out, but like others have said, the idea is not to make it impossible. The idea is to make it more trouble than it is worth.
I would use the https protocol with client-side keys too. You can use one client key for everyone or you can even generate a different key for each client and "register" them at your server.
I suppose that it's a lot of work for small project, but it sounds like the appropriate thing to do if you need authentication.
You should check that keys aren't seen easily by mobile phone owner. And remember that somebody will be able to hack it in any case.
Here's one thought - send up the device ID along with requests from your app.
Monitor the device ID's used - if you see a ton of requests from different IP's near or at the same time, that device is probably being used as a fixed key in the requests sent to you - block it.
For those that actually send the real device ID from other apps (not yours), you can monitor usage trends to see if the calls match the pattern of how your app performs - like one call being used by a device before some initialization call you would normally expect, and so on - block those too.
Basically by being able to shift rules around patterns of use, you can better adjust to someone trying to use your service by making sure it's not a fixed target like some random use key would be.
You may also want to use a simple use key as well as a first line of defense, and then layer on the traffic analysis approach. Also custom http header values you look for are another simple way to trip up a naive attacker.
I am assuming you don't want to use SSL? If you do then you can open HTTPS session and then pass some secret key in the request.
If you don't want SSL your options are limited: to have pseudo security I suggest both authentication and authorization methods and a third to reduce overall traffic:
Authentication: Generator in client application that creates secret keys by combining with a key file. The keyfile can be updated every so often for greater security: lets say you update the key file once a week. To re-cap: Generator combines in app secret with out of app key file to generate a 3rd key for transmission used in authentication. The server would then be able to authenticate.
Authorization: Of course you also want to lock out rogue applications. Here it would be best to have authorization mechanism with the site. Don't replace keyfiles for unless the client logs in. Track key files to users. etc.
Traffic reduction:
If you are receiving obscene amount of traffic or if you suspect someone trying to DOS your server, you can also have both the server and clients sync to request/response on a procedurally generated URL that can change often. It is wasteful to open/close so many HTTPS sessions if someone is just flooding you with requests.
I'm not sure what web technology you are using, but if you are using Ruby on Rails, it uses a secret authentication token in all of its controllers to make sure malicious code isn't accessing destructive methods (via PUSH, POST, or DELETE). You would need to send that authentication token to the server in your request body to allow it to execute. That should achieve what I think you are looking for.
If you're not using Ruby on Rails, that method of code authentication might be a good one to research and implement yourself in whatever technology you are using.
Take a look at the Rails Security Guide, specifically section 3.1 (CSRF Countermeasures).
You could do something like encrypting the current time and IP address from the iPhone, and then decrypt it on the server. The downside is that you need the iPhone app to know the "secret" key so that only it can generate valid access tokens... and once the key is in the wild, it will only be a matter of time before it's hacked if your app is really worth the effort.
You could encrypt the response using some random portion of the application which is meant to be using it, specifying the location of the binary in an unencrypted bit of the response. Then at least only clients with access to your binary would be able to decrypt it... but again, that's hardly 100% secure.
Ultimately you need to ask yourself how much effort you want to put into securing the service vs how much effort you think hackers will put into abusing it.
I'm not an Cocoa Touch developer, but I think HTTP Authentication over SSL would be easy to implement and it's probably exactly what you're looking for.
All you need to do is setup HTTP Authentication on the server side (you haven't mentioned what you're using on the server side) and create a self-signed SSL cert on your webserver. Done. :)
Tell us more about your setup and we will be able to help you further.
As some of the answers have stated, closing your web service off to everyone else will be a major hassle. The best you can hope for is to make it easier for the hackers to use another web service, than to use yours...
Another suggestion to do this, is to generate random numbers from a random seed on both the server and the client. The server would need to track where in the sequence of random numbers all of the clients are, and match that number to the one sent by the client.
The client would also have to register to be given access to the server. This would also serve as a authentication mechanism.
So:
//Client code:
$sequence = file_get_contents('sequence.txt');
$seed = file_get_contents('seed.txt');
$sequence++;
//Generate the $sequence-th random number
srand($seed);
for ($i = 0; $i <= $sequence; $i++) {
$num = rand();
}
//custom fetch function
get_info($main_url . '?num=' . $num . '&id' = $my_id);
This will generate a request similiar to this:
http://webservice.com/get_info.php?num=3489347&id=3
//Server Code: (I'm used to PHP)
//Get the ID and the random number
$id = (int)$_REQUEST['id'];
$rand = (int)$_REQUEST['num'];
$stmt = $db->prepare('SELECT `sequence`, `seed` FROM `client_list` WHERE `id` = :id');
if ($stmt->execute(array(':id' => $id)) {
list($sequence, $seed) = $stmt->fetch(PDO::FETCH_ASSOC);
}
$sequence++;
//Generate the $sequence-th random number
srand($seed);
for ($i = 0; $i <= $sequence; $i++) {
$num = rand();
}
if ($num == $rand) {
//Allow Access
} else {
//Deny Access
}
By using a a different seed for each client, you ensure that hackers can't predict a random number by tracking previous numbers used.

Store an encryption key in Keychain while application installation process

I need my application to use client's phone-number to generate unique ID for my web-service. Of course a phone-number is unique, but it must be secured. So it can be implemented with symmetric encryption (asymmetric will be later, because leak of resources), but I do not know where to store a encryption-key.
1.
I do not know why, but seems bad to store a key as a static field in code. May be because it's too easy to read it from here even not running an application.
2.
It seems better to store a key in Keychain and get it from here by request. But to avoid #1 it's necessary to install a key to Keychain while installation process. Is it possible? How to do that?
3.
I do not know what certificates do. Are they helpful to the problem?
4.
To transfer a key from server is also a bad idea, because it's very easy to sniffer it.
The way you solve the sniffing problem is that you communicate over HTTPS for your web service. NSURLConnection will do this easily, and all web service engines I know of handle HTTPS without trouble. This will get rid of many of your problems right away.
On which machine is the 100-1000x decrypt the bottleneck? Is your server so busy that it can't do an asym decryption? You should be doing this so infrequently on the phone that it should be irrelevant. I'm not saying asym is the answer here; only that its performance overhead shouldn't be the issue for securing a single string, decrypted once.
Your service requires SMS such that all users must provide their phone number? Are you trying to automate grabbing the phone number, or do you let the user enter it themselves? Automatically grabbing the phone number through the private APIs (or the non-private but undocumented configuration data) and sending that to a server is likely to run afoul of terms of service. This is a specific use-case Apple wants to protect the user from. You definitely need to be very clear in your UI that you are doing this and get explicit user permission.
Personally I'd authenticate as follows:
Server sends challenge byte
Client sends UUID, date, and hash(UUID+challenge+userPassword+obfuscationKey+date).
Server calculates same, makes sure date is in legal range (30-60s is good) and validates.
At this point I generally have the server generate a long, sparse, random session id which the client may use for the remainder of this "session" (anywhere from the next few minutes to the next year) rather than re-authenticating in every message.
ObfuscationKey is a secret key you hardcode into your program and server to make it harder for third parties to create bogus clients. It is not possible, period, not possible, to securely ensure that only your client can talk to your server. The obfuscationKey helps, however, especially on iPhone where reverse engineering is more difficult. Using UUID also helps because it is much less known to third-parties than phone number.
Note "userPassword" in there. The user should authenticate using something only the user knows. Neither the UUID nor the phone number is such a thing.
The system above, plus HTTPS, should be straightforward to implement (I've done it many times in many languages), have good performance, and be secure to an appropriate level for a broad range of "appropriate."
I don't think you're going to be able to do what you want securely with symmetric encryption. With asym you can send the public key without worrying about it too much (only threat is someone substituting your key with their own) and validate the encrypted unique id on your server with the private key.

How would you keep secret data secret in an iPhone application?

Let's say I need to access a web service from an iPhone app. This web service requires clients to digitally sign HTTP requests in order to prove that the app "knows" a shared secret; a client key. The request signature is stored in a HTTP header and the request is simply sent over HTTP (not HTTPS).
This key must stay secret at all times yet needs to be used by the iPhone app.
So, how would you securely store this key given that you've always been told to never store anything sensitive on the client side?
The average user (99% of users) will happily just use the application. There will be somebody (an enemy?) who wants that secret client key so as to do the service or client key owner harm by way of impersonation. Such a person might jailbreak their phone, get access to the binary, run 'strings' or a hex editor and poke around. Thus, just storing the key in the source code is a terrible idea.
Another idea is storing the key in code not a string literal but in a NSMutableArray that's created from byte literals.
One can use the Keychain but since an iPhone app never has to supply a password to store things in the Keychain, I'm wary that someone with access to the app's sandbox can and will be able to simply look at or trivially decode items therein.
EDIT - so I read this about the Keychain: "In iPhone OS, an application always has access to its own keychain items and does not have access to any other application’s items. The system generates its own password for the keychain, and stores the key on the device in such a way that it is not accessible to any application."
So perhaps this is the best place to store the key.... If so, how do I ship with the key pre-entered into the app's keychain? Is that possible? Else, how could you add the key on first launch without the key being in the source code? Hmm..
EDIT - Filed bug report # 6584858 at http://bugreport.apple.com
Thanks.
The goal is, ultimately, restrict access of the web service to authorized users, right? Very easy if you control the web service (if you don't -- wrap it in a web service which you do control).
1) Create a public/private key pair. The private key goes on the web service server, which is put in a dungeon and guarded by a dragon. The public key goes on the phone. If someone is able to read the public key, this is not a problem.
2) Have each copy of the application generate a unique identifier. How you do this is up to you. For example, you could build it into the executable on download (is this possible for iPhone apps)? You could use the phone's GUID, assuming they have a way of calculating one. You could also redo this per session if you really wanted.
3) Use the public key to encrypt "My unique identifier is $FOO and I approved this message". Submit that with every request to the web service.
4) The web service decrypts each request, bouncing any which don't contain a valid identifier. You can do as much or as little work as you want here: keep a whitelist/blacklist, monitor usage on a per-identifier basis and investigate suspicious behavior, etc.
5) Since the unique identifier now never gets sent over the wire, the only way to compromise it is to have physical access to the phone. If they have physical access to the phone, you lose control of any data anywhere on the phone. Always. Can't be helped. That is why we built the system such that compromising one phone never compromises more than one account.
6) Build business processes to accommodate the need to a) remove access from a user who is abusing it and b) restore access to a user whose phone has been physically compromised (this is going to be very, very infrequent unless the user is the adversary).
The simple answer is that as things stand today it's just not possible to keep secrets on the iPhone. A jailbroken iPhone is just a general-purpose computer that fits in your hand. There's no trusted platform hardware that you can access. The user can spoof anything you can imagine using to uniquely identify a given device. The user can inject code into your process to do things like inspect the keychain. (Search for MobileSubstrate to see what I mean.) Sorry, you're screwed.
One ray of light in this situation is in app purchase receipts. If you sell an item in your app using in app purchase you get a receipt that's crypto signed and can be verified with Apple on demand. Even though you can't keep the receipt secret it can be traced (by Apple, not you) to a specific purchase, which might discourage pirates from sharing them. You can also throttle access to your server on a per-receipt basis to prevent your server resources from being drained by pirates.
UAObfuscatedString could be a solution to your problem. From the docs:
When you write code that has a string constant in it, this string is saved in the binary in clear text. A hacker could potentially discover exploits or change the string to affect your app's behavior. UAObfuscatedString only ever stores single characters in the binary, then combines them at runtime to produce your string. It is highly unlikely that these single letters will be discoverable in the binary as they will be interjected at random places in the compiled code. Thus, they appear to be randomized code to anyone trying to extract strings.
If you can bear to be iPhone OS 3.0-only, you may want to look at push notifications. I can't go into the specifics, but you can deliver a payload to Apple's servers along with the notification itself. When they accept the alert (or if your app is running), then some part of your code is called and the keychain item is stored. At this point, that is the only route to securely storing a secret on an iPhone that I can think of.
I had the same question and spent a lot of time poking around for an answer. The issue is a chicken and egg one: how to pre-poluate the keychain with data needed by your app.
In any case, I found a technique that at least will make it harder for a jailbreaker to uncover the information - they'll at least have to disassemble your code to find out what you did to mask the info:
String Obfuscation (if the link breaks search for "Obfuscate / Encrypt a String (NSString)")
Essentially the string is obfuscated before placed in the app, then you unobfuscate it using code.
Its better than doing nothing.
David
EDIT: I actually used this in an app. I put a base coding string into the info.plist, then did several operations on it in code - rot13, rotate/invert bytes, etc. The final processed string was used to decode the obfuscated string. Now, the three letter agencies could for sure break this - but at a huge cost of many hours decoding the binary.
I was going to say that this is the best technique I've come across, but I just read Kiran's post on UAObfuscatedString (different answer), which is a completely different way to obfuscate. It has the benefit of no strings saved anywhere in the app - each letter is turned into a method call. The selectors will show up as strings, so a hacker can quickly tell that your class used that technique though.
I think that this similar question, and my answer, may be relevant to your case too. In a nutshell, there was some talk of a trusted platform module being present in an iPhone. This would allow your service to trust an iPhone, even in the hands of an attacker. However, it looks like using the keychain is your best bet.
Did you consider/try the Push Notification suggestion, for initially transmitting the secret to the app & keychain? Or end up finding some other method to achieve this?
I'm going have my iphone app upload images to Amazon S3. Instead of putting the AWS credentials in the app, I am going to have the app phone home to my server for the URI and headers to use in the S3 upload request. My server will generate the S3 URI, proper signatures, etc. I can then implement a tighter, more specific security model on my app's webservice than AWS offers by itself and not give away my AWS keys to anyone with a jailbroken iphone.
But there still has to be some trust (credentials or otherwise) given to the app, and that trust can be stolen. All you can ever do is limit the damage done if someone jailbreaks an iphone and steals whatever credentials are in the app. The more powerful those credentials are, the worst things are. Ways to limit the power of credentials include:
avoid global credentials. make them per-user/application
avoid permanent credentials. make them temporary if possible
avoid global permissions. give them only the permissions they need. for instance, write permissions might be broken down into insert, overwrite, delete, write against resource group A or B, etc, and read could be broken into read named resources, read a list of all existing resources, read resource groups A or B, etc.
I would recommend creating a key at run time if possible. This way if the key were to get apprehended during a particular session, once the session ends, the key will be worthless. They could still apprehend the key from memory if they are smart enough, but it wouldn't matter since the key would become invalid after a period of time.
Sounds wonky. Would use HTTPS and maybe an encryption package to handle the key.
I think CommonCrypto is available for iPhone.
EDIT: Still sounds wonky. Why would anyone pass a secret key in an HTTP header? Anyone who traces your network traffic (via a logging wifi router, for instance) would see it.
There are well-established security methods for encrypting message traffic...why not use them rather than invent what is basically a trivially flawed system?
EDIT II: Ah, I see. I would go ahead and use the Keychain...I think it is intended for just these kinds of cases. I missed that you were generating the request using the key. Would still use HTTPS if I could though, since that way you don't risk people deducing your keygeneration scheme via inspection of enough signatures.