sftp: Allow user to optionally check server hosts key to add security

Based on Issue 4087
  https://github.com/rclone/rclone/issues/4087

Current behaviour is insecure.  If the user specifies this value then we
switch to validating the server hostkey and so can detect server changes
or MITM-type attacks.

backend/sftp/sftp.go

@@ -32,6 +32,7 @@ import (
 	"github.com/rclone/rclone/lib/readers"
 	sshagent "github.com/xanzy/ssh-agent"
 	"golang.org/x/crypto/ssh"
+	"golang.org/x/crypto/ssh/knownhosts"
 )
 
 const (
@@ -86,6 +87,16 @@ in the new OpenSSH format can't be used.`,
 			Help: `Optional path to public key file.
 
 Set this if you have a signed certificate you want to use for authentication.` + env.ShellExpandHelp,
+		}, {
+			Name: "known_hosts_file",
+			Help: `Optional path to known_hosts file.
+
+Set this value to enable server host key validation.` + env.ShellExpandHelp,
+			Advanced: true,
+			Examples: []fs.OptionExample{{
+				Value: "~/.ssh/known_hosts",
+				Help:  "Use OpenSSH's known_hosts file",
+			}},
 		}, {
 			Name: "key_use_agent",
 			Help: `When set forces the usage of the ssh-agent.
@@ -195,6 +206,7 @@ type Options struct {
 	KeyFile           string `config:"key_file"`
 	KeyFilePass       string `config:"key_file_pass"`
 	PubKeyFile        string `config:"pubkey_file"`
+	KnownHostsFile    string `config:"known_hosts_file"`
 	KeyUseAgent       bool   `config:"key_use_agent"`
 	UseInsecureCipher bool   `config:"use_insecure_cipher"`
 	DisableHashCheck  bool   `config:"disable_hashcheck"`
@@ -414,6 +426,7 @@ func NewFs(name, root string, m configmap.Mapper) (fs.Fs, error) {
 	if opt.Port == "" {
 		opt.Port = "22"
 	}
+
 	sshConfig := &ssh.ClientConfig{
 		User:            opt.User,
 		Auth:            []ssh.AuthMethod{},
@@ -422,6 +435,14 @@ func NewFs(name, root string, m configmap.Mapper) (fs.Fs, error) {
 		ClientVersion:   "SSH-2.0-" + fs.Config.UserAgent,
 	}
 
+	if opt.KnownHostsFile != "" {
+		hostcallback, err := knownhosts.New(opt.KnownHostsFile)
+		if err != nil {
+			return nil, errors.Wrap(err, "couldn't parse known_hosts_file")
+		}
+		sshConfig.HostKeyCallback = hostcallback
+	}
+
 	if opt.UseInsecureCipher {
 		sshConfig.Config.SetDefaults()
 		sshConfig.Config.Ciphers = append(sshConfig.Config.Ciphers, "aes128-cbc", "aes192-cbc", "aes256-cbc", "3des-cbc")

docs/content/sftp.md

@@ -148,6 +148,57 @@ Note: the cert must come first in the file.  e.g.
 cat id_rsa-cert.pub id_rsa > merged_key
 ```
 
+### Host key validation ###
+
+By default rclone will not check the server's host key for validation.  This
+can allow an attacker to replace a server with their own and if you use
+password authentication then this can lead to that password being exposed.
+
+Host key matching, using standard `known_hosts` files can be turned on by
+enabling the `known_hosts_file` option.  This can point to the file maintained
+by `OpenSSH` or can point to a unique file.
+
+e.g.
+
+```
+[remote]
+type = sftp
+host = example.com
+user = sftpuser
+pass = 
+known_hosts_file = ~/.ssh/known_hosts
+````
+
+There are some limitations:
+
+* `rclone` will not _manage_ this file for you.  If the key is missing or
+wrong then the connection will be refused.
+* If the server is set up for a certificate host key then the entry in
+the `known_hosts` file _must_ be the `@cert-authority` entry for the CA
+* Unlike `OpenSSH`, the libraries used by `rclone` do not permit (at time
+of writing) multiple host keys to be listed for a server.  Only the first
+entry is used.
+
+If the host key provided by the server does not match the one in the
+file (or is missing) then the connection will be aborted and an error
+returned such as
+
+    NewFs: couldn't connect SSH: ssh: handshake failed: knownhosts: key mismatch
+
+or
+
+    NewFs: couldn't connect SSH: ssh: handshake failed: knownhosts: key is unknown
+
+If you see an error such as
+
+    NewFs: couldn't connect SSH: ssh: handshake failed: ssh: no authorities for hostname: example.com:22
+
+then it is likely the server has presented a CA signed host certificate
+and you will need to add the appropriate `@cert-authority` entry.
+
+The `known_hosts_file` setting can be set during `rclone config` as an
+advanced option.
+
 ### ssh-agent on macOS ###
 
 Note that there seem to be various problems with using an ssh-agent on
@@ -320,6 +371,23 @@ Leave blank or set to false to enable hashing (recommended), set to true to disa
 
 Here are the advanced options specific to sftp (SSH/SFTP Connection).
 
+#### --sftp-known-hosts-file
+
+Optional path to known_hosts file.
+
+Set this value to enable server host key validation.
+
+Leading `~` will be expanded in the file name as will environment variables such as `${RCLONE_CONFIG_DIR}`.
+
+
+- Config:      known_hosts_file
+- Env Var:     RCLONE_SFTP_KNOWN_HOSTS_FILE
+- Type:        string
+- Default:     ""
+- Examples:
+    - "~/.ssh/known_hosts"
+        - Use OpenSSH's known_hosts file
+
 #### --sftp-ask-password
 
 Allow asking for SFTP password when needed.