We would be creating N NWListener objects and M NWConnection objects in our process' communication subsystem to create server sockets, accepted client sockets on server and client sockets on clients. Both NWConnection and NWListener rely on DispatchQueue to deliver state changes, incoming connections, send/recv completions etc. What DispatchQueues should I use and why? Global Concurrent Dispatch Queue (and which QoS?) for all NWConnection and NWListener One custom concurrent queue (which QoS?) for all NWConnection and NWListener? (Does that anyways get targetted to one of the global queues?) One custom concurrent queue per NWConnection and NWListener though all targetted to Global Concurrent Dispatch Queue (and which QoS?)? One custom concurrent queue per NWConnection and NWListener though all targetted to single target custom concurrent queue? For every option above, how am I impacted in terms of parallelism, concurrency, throughput & latency and how is overall system impacted (with other processes also running)? Seperate questions (sorry for the digression): Are global concurrent queues specific to a process or shared across all processes on a device? Can I safely use setSpecific on global dispatch queues in our app?

I want to configure one aspect of my networking configuration (the QUIC keepalive interval). This only seems to be configurable via Network.framework’s nw_quic_set_keepalive_interval. Is there any way to apply this to a URLSession? Or do I need to implement the whole connection management myself using Network.framework?

When setting new entitlements com.apple.developer.networking.carrier-constrained.appcategory and com.apple.developer.networking.carrier-constrained.app-optimized, I have a question about how URLSession should behave. I notice we have a way to specify whether a Network connection should allow ultra-constrained paths via NWParameters allowUltraConstrainedPaths: https://developer.apple.com/documentation/network/nwparameters/allowultraconstrainedpaths There does not appear to be a similar property on URLSessionConfiguration. In an ultra-constrained (eg. satellite) network, should we expect all requests made through an URLSession to fail? Does all network activity when ultra-constrained need to go through a NWConnection or NetworkConnection specifically configured with allowUltraConstrainedPaths, or can URLSession ever be configured to allow ultra-constrained paths?

Hello, I am working to integrate the new com.apple.developer.networking.carrier-constrained.app-optimized entitlement in my iOS 26 app so that my app can use a carrier-provided satellite network, and want to confirm my understanding of how to detect and optimize for satellite network conditions. (Ref: https://developer.apple.com/documentation/bundleresources/entitlements/com.apple.developer.networking.carrier-constrained.app-optimized ) My current approach: I plan to set the entitlement to true once my app is optimized for satellite networks. To detect if the device is connected to a satellite network, I intend to use the Network framework’s NWPath properties: isUltraConstrained — I understand this should be set to true when the device is connected to a satellite network. (Ref: https://developer.apple.com/documentation/network/nwpath/isultraconstrained ) linkQuality == .minimal — I believe this will also be set in satellite scenarios, though it may not be exclusive to satellite connections. (Ref: https://developer.apple.com/documentation/network/nwpath/linkquality-swift.enum/minimal ) Questions: Is it correct that isUltraConstrained will reliably indicate a satellite connection? Should I also check for linkQuality == .minimal, or is isUltraConstrained sufficient? Are there any additional APIs or best practices for detecting and optimizing for satellite connectivity that I should be aware of? Thank you for confirming whether my understanding and approach are correct, and for any additional guidance.

Hi, I am in need of your help with publishing my game. I got the following explanation for the negative review of my app/game. Issue Description One or more purpose strings in the app do not sufficiently explain the use of protected resources. Purpose strings must clearly and completely describe the app's use of data and, in most cases, provide an example of how the data will be used. Next Steps Update the local network information purpose string to explain how the app will use the requested information and provide a specific example of how the data will be used. See the attached screenshot. Resources Purpose strings must clearly describe how an app uses the ability, data, or resource. The following are hypothetical examples of unclear purpose strings that would not pass review: "App would like to access your Contacts" "App needs microphone access" See examples of helpful, informative purpose strings. The problem is that they say my app asks to allow my app to find devices on local networks. And that this needs more explanation in the purpose strings. Totally valid to ask, but the problem is my app doesn't need local access to devices, and there shouldn't be code that asks this?? FYI the game is build with Unity. Would love some help on how to turn this off so that my app can get published.

I want to know the reason for last Wi-Fi disconnection. It would be helpful to that I can get the reason-code received from the access point.

It works when one device is only a publisher and the other is only a subscriber. However, when both devices act as both publisher and subscriber simultaneously—which Apple’s documentation (https://developer.apple.com/documentation/wifiaware/adopting-wi-fi-aware#Declare-services) indicates is valid—the connection never establishes. After timing out, both NetworkListener and NetworkBrowser transition to the failed state. This appears to be a race condition in Network framework. Task.detached { try await NetworkListener( for: .wifiAware( .connecting( to: .myService, from: .allPairedDevices, datapath: .defaults ) ), using: .parameters { Coder( sending: ..., receiving: ..., using: NetworkJSONCoder() ) { TCP() } } ).run { connection in await self.add(connection: connection) } } Task.detached { try await NetworkBrowser( for: .wifiAware( .connecting( to: .allPairedDevices, from: .myService ) ), using: .tcp ).run { endpoints in for endpoint in endpoints { await self.connect(to: endpoint) } } }

NWPathMonitor appears to retain itself (or is retained by some internal infrastructure) once it has been started until cancelled. This seems like it can lead to memory leaks if the references to to the monitor are dropped. Is this behavior documented anywhere? func nwpm_self_retain() { weak var weakRef: NWPathMonitor? autoreleasepool { let monitor: NWPathMonitor = NWPathMonitor() weakRef = monitor monitor.start(queue: .main) // monitor.cancel() // assertion fails unless this is called } assert(weakRef == nil) } nwpm_self_retain()

I have been using the SCNetworkReachabilityGetFlags for 10+ years to inform users that their request won't work. In my experience this works pretty well although i am aware of the limitations. Now, i am looking into the NWPathMonitor, and i have one situation that i'm trying to. get my head around - it's asynchronous. Specifically, i am wondering what to do when my geofences trigger and i want to check network connectivity - i want to tell the user why the operation i'll perform because of the trigger couldn't be done. SO. say i start a NWPathMonitor in didFinishLaunchingWithOptions. When the app is booted up because of a geofence trigger, might i not end up in a case where my didEnterRegion / didExitRegion gets called before the NWPathMonitor has gotten its first status? The advantage here with SCNetworkReachabilityGetFlags, as i understand it, would be that it's synchronous? If i want to upgrade to nwpathmonitor, i guess i have to do a method that creates a nwpathmonitor, uses a semaphore to wait for the first callback, then contunues? Thoughts appreciated

Hello, I have an app that was using the iOS 18 Network Framework APIs. It used Peer to Peer, QUIC and Bonjour. It was all working as expected. I wanted to upgrade to the new iOS 26 Network Framework APIs (NetworkBrowser, NetworkListener, NetworkConnection...). I have things working (multiple devices can discover each other, connection to each other and send messages to each other) but my app crashes when I go to toggle of all the networking stuff. In the iOS 18 Network Framework API NWConnection had a .cancel() function I could use to tell the other side the connection was done. I dont see a cancel function for NetworkConnection. My question is - how do I properly close down a NetworkConnection and also properly tell the other side the connection is done.

For important background information, read Extra-ordinary Networking before reading this. Share and Enjoy — Quinn “The Eskimo!” @ Developer Technical Support @ Apple let myEmail = "eskimo" + "1" + "@" + "apple.com" Network Interface APIs Most developers don’t need to interact directly with network interfaces. If you do, read this post for a summary of the APIs available to you. Before you read this, read Network Interface Concepts. Interface List The standard way to get a list of interfaces and their addresses is getifaddrs. To learn more about this API, see its man page. A network interface has four fundamental attributes: A set of flags — These are packed into a CUnsignedInt. The flags bits are declared in <net/if.h>, starting with IFF_UP. An interface type — See Network Interface Type, below. An interface index — Valid indexes are greater than 0. A BSD interface name. For example, an Ethernet interface might be called en0. The interface name is shared between multiple network interfaces running over a given hardware interface. For example, IPv4 and IPv6 running over that Ethernet interface will both have the name en0. WARNING BSD interface names are not considered API. There’s no guarantee, for example, that an iPhone’s Wi-Fi interface is en0. You can map between the last two using if_indextoname and if_nametoindex. See the if_indextoname man page for details. An interface may also have address information. If present, this always includes the interface address (ifa_addr) and the network mask (ifa_netmask). In addition: Broadcast-capable interfaces (IFF_BROADCAST) have a broadcast address (ifa_broadaddr, which is an alias for ifa_dstaddr). Point-to-point interfaces (IFF_POINTOPOINT) have a destination address (ifa_dstaddr). Calling getifaddrs from Swift is a bit tricky. For an example of this, see QSocket: Interfaces. IP Address List Once you have getifaddrs working, it’s relatively easy to manipulate the results to build a list of just IP addresses, a list of IP addresses for each interface, and so on. QSocket: Interfaces has some Swift snippets that show this. Interface List Updates The interface list can change over time. Hardware interfaces can be added and removed, network interfaces come up and go down, and their addresses can change. It’s best to avoid caching information from getifaddrs. If thats unavoidable, use the kNotifySCNetworkChange Darwin notification to update your cache. For information about registering for Darwin notifications, see the notify man page (in section 3). This notification just tells you that something has changed. It’s up to you to fetch the new interface list and adjust your cache accordingly. You’ll find that this notification is sometimes posted numerous times in rapid succession. To avoid unnecessary thrashing, debounce it. While the Darwin notification API is easy to call from Swift, Swift does not import kNotifySCNetworkChange. To fix that, define that value yourself, calling a C function to get the value: var kNotifySCNetworkChange: UnsafePointer<CChar> { networkChangeNotifyKey() } Here’s what that C function looks like: extern const char * networkChangeNotifyKey(void) { return kNotifySCNetworkChange; } Network Interface Type There are two ways to think about a network interface’s type. Historically there were a wide variety of weird and wonderful types of network interfaces. The following code gets this legacy value for a specific BSD interface name: func legacyTypeForInterfaceNamed(_ name: String) -> UInt8? { var addrList: UnsafeMutablePointer<ifaddrs>? = nil let err = getifaddrs(&addrList) // In theory we could check `errno` here but, honestly, what are gonna // do with that info? guard err >= 0, let first = addrList else { return nil } defer { freeifaddrs(addrList) } return sequence(first: first, next: { $0.pointee.ifa_next }) .compactMap { addr in guard let nameC = addr.pointee.ifa_name, name == String(cString: nameC), let sa = addr.pointee.ifa_addr, sa.pointee.sa_family == AF_LINK, let data = addr.pointee.ifa_data else { return nil } return data.assumingMemoryBound(to: if_data.self).pointee.ifi_type } .first } The values are defined in <net/if_types.h>, starting with IFT_OTHER. However, this value is rarely useful because many interfaces ‘look like’ Ethernet and thus have a type of IFT_ETHER. Network framework has the concept of an interface’s functional type. This is an indication of how the interface fits into the system. There are two ways to get an interface’s functional type: If you’re using Network framework and have an NWInterface value, get the type property. If not, call ioctl with a SIOCGIFFUNCTIONALTYPE request. The return values are defined in <net/if.h>, starting with IFRTYPE_FUNCTIONAL_UNKNOWN. Swift does not import SIOCGIFFUNCTIONALTYPE, so it’s best to write this code in a C: extern uint32_t functionalTypeForInterfaceNamed(const char * name) { int fd = socket(AF_INET, SOCK_DGRAM, 0); if (fd < 0) { return IFRTYPE_FUNCTIONAL_UNKNOWN; } struct ifreq ifr = {}; strlcpy(ifr.ifr_name, name, sizeof(ifr.ifr_name)); bool success = ioctl(fd, SIOCGIFFUNCTIONALTYPE, &ifr) >= 0; int junk = close(fd); assert(junk == 0); if ( ! success ) { return IFRTYPE_FUNCTIONAL_UNKNOWN; } return ifr.ifr_ifru.ifru_functional_type; } Finally, TN3158 Resolving Xcode 15 device connection issues documents the SIOCGIFDIRECTLINK flag as a specific way to identify the network interfaces uses by Xcode for device connection traffic. Revision History 2025-12-10 Added info about SIOCGIFDIRECTLINK. 2023-07-19 First posted.

Summary NetworkConnection<WebSocket> in iOS 26 Network framework throws POSIXErrorCode(rawValue: 22): Invalid argument when receiving WebSocket ping (opcode 9) or pong (opcode 10) control frames. This prevents proper WebSocket keep-alive functionality. Environment iOS 26.0 (Simulator) macOS 26.1 Xcode 26.0 Note: This issue was initially discovered on iOS 26 Simulator. The same behavior was confirmed on macOS 26, suggesting a shared bug in the Network framework. The attached sample code is for macOS for easier reproduction. Description When using the new NetworkConnection<WebSocket> API introduced in iOS 26 or macOS 26, the receive() method throws EINVAL error whenever a ping or pong control frame is received from the server. This is a critical issue because: WebSocket servers commonly send ping frames to keep connections alive Clients send ping frames to verify connection health The receive callback never receives the ping/pong frame - the error occurs before the frame reaches user code Steps to Reproduce Create a WebSocket connection to any server that supports ping/pong (e.g., wss://echo.websocket.org): import Foundation import Network // MARK: - WebSocket Ping/Pong EINVAL Bug Reproduction // This sample demonstrates that NetworkConnection<WebSocket> throws EINVAL // when receiving ping or pong control frames. @main struct WebSocketPingPongBug { static func main() async { print("=== WebSocket Ping/Pong EINVAL Bug Reproduction ===\n") do { try await testPingPong() } catch { print("Test failed with error: \(error)") } } static func testPingPong() async throws { let host = "echo.websocket.org" let port: UInt16 = 443 print("Connecting to wss://\(host)...") let endpoint = NWEndpoint.hostPort( host: NWEndpoint.Host(host), port: NWEndpoint.Port(rawValue: port)! ) try await withNetworkConnection(to: endpoint, using: { WebSocket { TLS { TCP() } } }) { connection in print("Connected!\n") // Start receive loop in background let receiveTask = Task { var messageCount = 0 while !Task.isCancelled { do { let (data, metadata) = try await connection.receive() messageCount += 1 print("[\(messageCount)] Received frame - opcode: \(metadata.opcode)") if let text = String(data: data, encoding: .utf8) { print("[\(messageCount)] Content: \(text)") } else { print("[\(messageCount)] Binary data: \(data.count) bytes") } } catch let error as NWError { if case .posix(let code) = error, code == .EINVAL { print("❌ EINVAL error occurred! (POSIXErrorCode 22: Invalid argument)") print(" This is the bug - ping/pong frame caused EINVAL") // Continue to demonstrate workaround continue } print("Receive error: \(error)") break } catch { print("Receive error: \(error)") break } } } // Wait for initial message from server try await Task.sleep(for: .seconds(2)) // Test 1: Send text message (should work) print("\n--- Test 1: Sending text message ---") try await connection.send("Hello, WebSocket!") print("✅ Text message sent") try await Task.sleep(for: .seconds(1)) // Test 2: Send ping (pong response will cause EINVAL) print("\n--- Test 2: Sending ping frame ---") print("Expecting EINVAL when pong is received...") let pingMetadata = NWProtocolWebSocket.Metadata(opcode: .ping) try await connection.ping(Data()) { pingMetadata } print("✅ Ping sent, waiting for pong...") // Wait for pong response try await Task.sleep(for: .seconds(2)) // Cleanup receiveTask.cancel() print("\n=== Test Complete ===") print("If you saw 'EINVAL error occurred!' above, the bug is reproduced.") } } } The receive() call fails with error when pong arrives: ❌ EINVAL error occurred! (POSIXErrorCode 22: Invalid argument) Test Results Scenario Result Send/receive text (opcode 1) ✅ OK Client sends ping, receives pong ❌ EINVAL on pong receive Expected Behavior The receive() method should successfully return ping and pong frames, or at minimum, handle them internally without throwing an error. The autoReplyPing option should allow automatic pong responses without disrupting the receive loop. Actual Behavior When a ping or pong control frame is received: The receive() method throws NWError.posix(.EINVAL) The frame never reaches user code (no opcode check is possible) The connection remains valid, but the receive loop is interrupted Workaround Catch the EINVAL error and restart the receive loop: while !Task.isCancelled { do { let received = try await connection.receive() // Process message } catch let error as NWError { if case .posix(let code) = error, code == .EINVAL { // Control frame caused EINVAL, continue receiving continue } throw error } } This workaround allows continued operation but: Cannot distinguish between ping-related EINVAL and other EINVAL errors Cannot access the ping/pong frame content Cannot implement custom ping/pong handling Impact WebSocket connections to servers that send periodic pings will experience repeated EINVAL errors Applications must implement workarounds that may mask other legitimate errors Additional Information Packet capture confirms ping/pong frames are correctly transmitted at the network level The error occurs in the Network framework's internal processing, before reaching user code

Recent our APP performance online has revealed significant degradation in cellular network SRTT (Smoothed Round-Trip Time) on the latest iPhone models (iPhone 18.1, 18.2, and 18.3) relative to previous generation devices. IDC network transmission SRTT P50 increased by 10.64%, P95 increased by 103.41%; CDN network transmission SRTT P50 increased by 12.66%, P95 increased by 81.08%. Detailed Performance Metrics: 1. Network Transmission SRTT Degradation Following optimization of our APP's network library, iOS network transmission SRTT showed improvement from mid-August through mid-September. However, starting September 16, cellular network SRTT metrics began to degrade (SRTT increased). This degradation affects both IDC and CDN routes. WiFi network performance remains unaffected. 2. Excluding iOS 26.x Version Data After data filtering, we discovered that the increase in iOS cellular network transmission SRTT was caused by data samples from iOS 26.x versions. When excluding iOS 26.x version data, network transmission SRTT shows no growth. 3. Comparative Analysis: iOS 26.x vs. iOS < 26.0 network transmission SRTT shows： IDC (Internet Data Center) Links: P50 latency: 10.64% increase / P95 latency: 103.41% increase CDN (Content Delivery Network) Links: P50 latency: 12.66% increase / P95 latency: 81.08% increase 4. Device-Model Analysis: iOS 26.x SRTT Degradation Scope Granular analysis of iOS 26.x samples across different device models reveals that network SRTT degradation is not universal but rather specific to certain iPhone models. These measurements indicate a substantial regression in network performance across both data center and content delivery pathways.

We are facing a DNS resolution issue with a specific ISP, where our domain name does not resolve correctly using the system DNS. However, the same domain works as expected when a custom DNS resolver is used. On Android, this is straightforward to handle by configuring a custom DNS implementation using OkHttp / Retrofit. I am trying to implement a functionally equivalent solution in native iOS (Swift / SwiftUI). Android Reference (Working Behavior) : val dns = DnsOverHttps.Builder() .client(OkHttpClient()) .url("https://cloudflare-dns.com/dns-query".toHttpUrl()) .bootstrapDnsHosts(InetAddress.getByName("1.1.1.1")) .build() OkHttpClient.Builder() .dns(dns) .build() Attempted iOS Approach I attempted the following approach : Resolve the domain to an IP address programmatically (using DNS over HTTPS) Connect directly to the resolved IP address Set the original domain in the Host HTTP header DNS Resolution via DoH : func resolveDomain(domain: String) async throws -> String {     guard let url = URL(         string: "https://cloudflare-dns.com/dns-query?name=\(domain)&type=A"     ) else {         throw URLError(.badURL)     }     var request = URLRequest(url: url)     request.setValue("application/dns-json", forHTTPHeaderField: "accept")     let (data, _) = try await URLSession.shared.data(for: request)     let response = try JSONDecoder().decode(DNSResponse.self, from: data)     guard let ip = response.Answer?.first?.data else {         throw URLError(.cannotFindHost)     }     return ip } API Call Using Resolved IP :  func callAPIUsingCustomDNS() async throws {     let ip = try await resolveDomain(domain: "example.com")     guard let url = URL(string: "https://(ip)") else {         throw URLError(.badURL)     }     let configuration = URLSessionConfiguration.ephemeral     let session = URLSession(         configuration: configuration,         delegate: CustomURLSessionDelegate(originalHost: "example.com"),         delegateQueue: .main     )     var request = URLRequest(url: url)     request.setValue("example.com", forHTTPHeaderField: "Host")     let (_, response) = try await session.data(for: request)     print("Success: (response)") } Problem Encountered When connecting via the IP address, the TLS handshake fails with the following error: Error Domain=NSURLErrorDomain Code=-1200 "A TLS error caused the secure connection to fail." This appears to happen because iOS sends the IP address as the Server Name Indication (SNI) during the TLS handshake, while the server’s certificate is issued for the domain name. Custom URLSessionDelegate Attempt :  class CustomURLSessionDelegate: NSObject, URLSessionDelegate {     let originalHost: String     init(originalHost: String) {         self.originalHost = originalHost     }     func urlSession(         _ session: URLSession,         didReceive challenge: URLAuthenticationChallenge,         completionHandler: @escaping (URLSession.AuthChallengeDisposition, URLCredential?) -> Void     ) {         guard challenge.protectionSpace.authenticationMethod == NSURLAuthenticationMethodServerTrust,               let serverTrust = challenge.protectionSpace.serverTrust else {             completionHandler(.performDefaultHandling, nil)             return         }         let sslPolicy = SecPolicyCreateSSL(true, originalHost as CFString)         let basicPolicy = SecPolicyCreateBasicX509()         SecTrustSetPolicies(serverTrust, [sslPolicy, basicPolicy] as CFArray)         var error: CFError?         if SecTrustEvaluateWithError(serverTrust, &error) {             completionHandler(.useCredential, URLCredential(trust: serverTrust))         } else {             completionHandler(.cancelAuthenticationChallenge, nil)         }     } } However, TLS validation still fails because the SNI remains the IP address, not the domain. I would appreciate guidance on the supported and App Store–compliant way to handle ISP-specific DNS resolution issues on iOS. If custom DNS or SNI configuration is not supported, what alternative architectural approaches are recommended by Apple?

I want to add more cipher suites. I use NWConnection to make a connection. Before I use sec_protocol_options_append_tls_ciphersuite method to add more cipher suites, I found that Apple provided 20 cipher suites shown in the client hello packet. But after I added three more cipher suites, I found that nothing changed, and still original 20 cipher suites shown in the client hello packet when I made a new connection. The following is the code about connection. I want to add three more cipher suites: tls_ciphersuite_t.ECDHE_ECDSA_WITH_AES_128_CBC_SHA256, tls_ciphersuite_t.ECDHE_ECDSA_WITH_AES_256_CBC_SHA384, tls_ciphersuite_t.ECDHE_RSA_WITH_AES_256_CBC_SHA384 Can you give me some advice about how to add more cipher suites? Thanks. By the way, I working on a MacOS app. Xcode version: 16 MacOS version: 15.6

Hi, I’m using Network Framework to implement a UDP client via NWConnection, and I’m looking for clarification about the correct and fully safe shutdown procedure, especially regarding resource release. I have initiated some pending receive calls on the NWConnection (using receive). After calling connection.cancel(), do we need to wait for the cancellation of these pending receives? As mentioned in this thread, NWConnection retains references to the receive closures and releases them once they are called. If a receive closure holds a reference to the NWConnection itself, do we need to wait for these closures to be called to avoid memory leaks? Or, if there are no such retained references, we don't need to wait for the cancellation of the pending I/O and cancelled state for NWConnection?

Hi, I’m using Network Framework to implement a UDP listener via NWListener. I am looking for clarification about the correct and fully safe shutdown procedure, especially regarding resource release. After calling listener.cancel(), do we need to wait for the .cancelled state before exiting the application? Or can we just exit once the cancellation is initiated, assuming the OS will close the NWListener and there will be no resource leak? Is it recommended (or required) to set newConnectionHandler = nil when shutting down a UDP listener? My understanding is that if there is no NWListener attached, then whenever a connection is accepted by the OS, it will not be delivered to the application and the OS will simply drop it.

Hi, I played around the last days with the new NetworkConnection API from Network framework that supports structured concurrency. I discovered a behavior, which is unexpected from my understanding. Let's say you have a dead endpoint or something that does not exist. Something where you receive a noSuchRecord error. When I then try to send data, I would expect that the send function throws an error but this does not happen. The function now suspends indefinitely which is well not a great behavior. Example simplified: func send() async { let connection = NetworkConnection(to: .hostPort(host: "apple.co.com", port: 8080)) { TCP() } do { try await connection.send("Hello World!".raw) } catch { print(error) } } I'm not sure if this is the intended behavior or how this should be handled. Thanks and best regards, Vinz

Hello, I am studying the Building peer-to-peer apps codebase https://developer.apple.com/documentation/wifiaware/building-peer-to-peer-apps and am wondering why no connection is ever started? I searched the codebase and didn't find .start() be called once. Start function I'm referencing https://developer.apple.com/documentation/network/networkconnection/start() Are NetworkConnections started automatically? Note that I am using QUIC NetworkConnections (NetworkConnection) in what I'm trying to do.

Transport Layer Security (TLS) is the most important security protocol on the Internet today. Most notably, TLS puts the S into HTTPS, adding security to the otherwise insecure HTTP protocol. IMPORTANT TLS is the successor to the Secure Sockets Layer (SSL) protocol. SSL is no longer considered secure and it’s now rarely used in practice, although many folks still say SSL when they mean TLS. TLS is a complex protocol. Much of that complexity is hidden from app developers but there are places where it’s important to understand specific details of the protocol in order to meet your requirements. This post explains the fundamentals of TLS, concentrating on the issues that most often confuse app developers. Note The focus of this is TLS-PKI, where PKI stands for public key infrastructure. This is the standard TLS as deployed on the wider Internet. There’s another flavour of TLS, TLS-PSK, where PSK stands for pre-shared key. This has a variety of uses, but an Apple platforms we most commonly see it with local traffic, for example, to talk to a Wi-Fi based accessory. For more on how to use TLS, both TLS-PKI and TLS-PSK, in a local context, see TLS For Accessory Developers. Server Certificates For standard TLS to work the server must have a digital identity, that is, the combination of a certificate and the private key matching the public key embedded in that certificate. TLS Crypto Magic™ ensures that: The client gets a copy of the server’s certificate. The client knows that the server holds the private key matching the public key in that certificate. In a typical TLS handshake the server passes the client a list of certificates, where item 0 is the server’s certificate (the leaf certificate), item N is (optionally) the certificate of the certificate authority that ultimately issued that certificate (the root certificate), and items 1 through N-1 are any intermediate certificates required to build a cryptographic chain of trust from 0 to N. Note The cryptographic chain of trust is established by means of digital signatures. Certificate X in the chain is issued by certificate X+1. The owner of certificate X+1 uses their private key to digitally sign certificate X. The client verifies this signature using the public key embedded in certificate X+1. Eventually this chain terminates in a trusted anchor, that is, a certificate that the client trusts by default. Typically this anchor is a self-signed root certificate from a certificate authority. Note Item N is optional for reasons I’ll explain below. Also, the list of intermediate certificates may be empty (in the case where the root certificate directly issued the leaf certificate) but that’s uncommon for servers in the real world. Once the client gets the server’s certificate, it evaluates trust on that certificate to confirm that it’s talking to the right server. There are three levels of trust evaluation here: Basic X.509 trust evaluation checks that there’s a cryptographic chain of trust from the leaf through the intermediates to a trusted root certificate. The client has a set of trusted root certificates built in (these are from well-known certificate authorities, or CAs), and a site admin can add more via a configuration profile. This step also checks that none of the certificates have expired, and various other more technical criteria (like the Basic Constraints extension). Note This explains why the server does not have to include the root certificate in the list of certificates it passes to the client; the client has to have the root certificate installed if trust evaluation is to succeed. In addition, TLS trust evaluation (per RFC 2818) checks that the DNS name that you connected to matches the DNS name in the certificate. Specifically, the DNS name must be listed in the Subject Alternative Name extension. Note The Subject Alternative Name extension can also contain IP addresses, although that’s a much less well-trodden path. Also, historically it was common to accept DNS names in the Common Name element of the Subject but that is no longer the case on Apple platforms. App Transport Security (ATS) adds its own security checks. Basic X.509 and TLS trust evaluation are done for all TLS connections. ATS is only done on TLS connections made by URLSession and things layered on top URLSession (like WKWebView). In many situations you can override trust evaluation; for details, see Technote 2232 HTTPS Server Trust Evaluation). Such overrides can either tighten or loosen security. For example: You might tighten security by checking that the server certificate was issued by a specific CA. That way, if someone manages to convince a poorly-managed CA to issue them a certificate for your server, you can detect that and fail. You might loosen security by adding your own CA’s root certificate as a trusted anchor. IMPORTANT If you rely on loosened security you have to disable ATS. If you leave ATS enabled, it requires that the default server trust evaluation succeeds regardless of any customisations you do. Mutual TLS The previous section discusses server trust evaluation, which is required for all standard TLS connections. That process describes how the client decides whether to trust the server. Mutual TLS (mTLS) is the opposite of that, that is, it’s the process by which the server decides whether to trust the client. Note mTLS is commonly called client certificate authentication. I avoid that term because of the ongoing industry-wide confusion between certificates and digital identities. While it’s true that, in mTLS, the server authenticates the client certificate, to set this up on the client you need a digital identity, not a certificate. mTLS authentication is optional. The server must request a certificate from the client and the client may choose to supply one or not (although if the server requests a certificate and the client doesn’t supply one it’s likely that the server will then fail the connection). At the TLS protocol level this works much like it does with the server certificate. For the client to provide this certificate it must apply a digital identity, known as the client identity, to the connection. TLS Crypto Magic™ assures the server that, if it gets a certificate from the client, the client holds the private key associated with that certificate. Where things diverge is in trust evaluation. Trust evaluation of the client certificate is done on the server, and the server uses its own rules to decided whether to trust a specific client certificate. For example: Some servers do basic X.509 trust evaluation and then check that the chain of trust leads to one specific root certificate; that is, a client is trusted if it holds a digital identity whose certificate was issued by a specific CA. Some servers just check the certificate against a list of known trusted client certificates. When the client sends its certificate to the server it actually sends a list of certificates, much as I’ve described above for the server’s certificates. In many cases the client only needs to send item 0, that is, its leaf certificate. That’s because: The server already has the intermediate certificates required to build a chain of trust from that leaf to its root. There’s no point sending the root, as I discussed above in the context of server trust evaluation. However, there are no hard and fast rules here; the server does its client trust evaluation using its own internal logic, and it’s possible that this logic might require the client to present intermediates, or indeed present the root certificate even though it’s typically redundant. If you have problems with this, you’ll have to ask the folks running the server to explain its requirements. Note If you need to send additional certificates to the server, pass them to the certificates parameter of the method you use to create your URLCredential (typically init(identity:certificates:persistence:)). One thing that bears repeating is that trust evaluation of the client certificate is done on the server, not the client. The client doesn’t care whether the client certificate is trusted or not. Rather, it simply passes that certificate the server and it’s up to the server to make that decision. When a server requests a certificate from the client, it may supply a list of acceptable certificate authorities [1]. Safari uses this to filter the list of client identities it presents to the user. If you are building an HTTPS server and find that Safari doesn’t show the expected client identity, make sure you have this configured correctly. If you’re building an iOS app and want to implement a filter like Safari’s, get this list using: The distinguishedNames property, if you’re using URLSession The sec_protocol_metadata_access_distinguished_names routine, if you’re using Network framework [1] See the certificate_authorities field in Section 7.4.4 of RFC 5246, and equivalent features in other TLS versions. Self-Signed Certificates Self-signed certificates are an ongoing source of problems with TLS. There’s only one unequivocally correct place to use a self-signed certificate: the trusted anchor provided by a certificate authority. One place where a self-signed certificate might make sense is in a local environment, that is, securing a connection between peers without any centralised infrastructure. However, depending on the specific circumstances there may be a better option. TLS For Accessory Developers discusses this topic in detail. Finally, it’s common for folks to use self-signed certificates for testing. I’m not a fan of that approach. Rather, I recommend the approach described in QA1948 HTTPS and Test Servers. For advice on how to set that up using just your Mac, see TN2326 Creating Certificates for TLS Testing. TLS Standards RFC 6101 The Secure Sockets Layer (SSL) Protocol Version 3.0 (historic) RFC 2246 The TLS Protocol Version 1.0 RFC 4346 The Transport Layer Security (TLS) Protocol Version 1.1 RFC 5246 The Transport Layer Security (TLS) Protocol Version 1.2 RFC 8446 The Transport Layer Security (TLS) Protocol Version 1.3 RFC 4347 Datagram Transport Layer Security RFC 6347 Datagram Transport Layer Security Version 1.2 RFC 9147 The Datagram Transport Layer Security (DTLS) Protocol Version 1.3 Share and Enjoy — Quinn “The Eskimo!” @ Developer Technical Support @ Apple let myEmail = "eskimo" + "1" + "@" + "apple.com" Revision History: 2025-11-21 Clearly defined the terms TLS-PKI and TLS-PSK. 2024-03-19 Adopted the term mutual TLS in preference to client certificate authentication throughout, because the latter feeds into the ongoing certificate versus digital identity confusion. Defined the term client identity. Added the Self-Signed Certificates section. Made other minor editorial changes. 2023-02-28 Added an explanation mTLS acceptable certificate authorities. 2022-12-02 Added links to the DTLS RFCs. 2022-08-24 Added links to the TLS RFCs. Made other minor editorial changes. 2022-06-03 Added a link to TLS For Accessory Developers. 2021-02-26 Fixed the formatting. Clarified that ATS only applies to URLSession. Minor editorial changes. 2020-04-17 Updated the discussion of Subject Alternative Name to account for changes in the 2019 OS releases. Minor editorial updates. 2018-10-29 Minor editorial updates. 2016-11-11 First posted.