skip to content

RDP

RDP multiplexes a Windows desktop into virtual channels, authenticates with NLA before a session exists and runs over TLS on 3389. Interviews focus on exposure: pre-auth bugs and brute force.

on this pageshow

explore

questions

30

In RDP, how does Standard RDP Security differ from TLS and CredSSP, and why is it the layer to retire?

level: juniorimportance: must knowfreq 64%

answer

  1. three layers, one is native
  2. RC4 keys from two randoms
  3. who authenticates the server?
  4. external protocol does the work
  5. legacy kept for old clients

basics

~20 s

Standard RDP Security is RDP's own legacy layer: RC4 session keys and no server authentication in the sequence itself. TLS and CredSSP hand protection and server authentication to an external protocol, so the legacy layer is the one to retire.

solid answer

~40 s

RDP can secure a session three ways. Standard RDP Security (`PROTOCOL_RDP`, 0x00000000) is RDP's native layer: RC4 session keys (40, 56 or 128 bit) built from two 32-byte random values, or FIPS-validated methods at the FIPS level. TLS (`PROTOCOL_SSL`) and CredSSP (`PROTOCOL_HYBRID`, which runs a TLS handshake first and then authenticates the user inside it) are Enhanced RDP Security: an external protocol does the encryption, the integrity checks and the server authentication. The specification states that the base connection sequence has no mechanism that authenticates the server, and its own glossary notes RC4's weaknesses. The legacy layer survives for compatibility with older clients; the control is to require TLS or CredSSP on the server and plan the retirement of clients that cannot offer them.

go deeper

for a junior

Name the three layers, say that Standard RDP Security is RDP's own RC4-based layer, and say TLS and CredSSP rely on an external protocol.

for a middle

Explain where the session keys come from in each layer, that the server's method and level are zero under Enhanced RDP Security, and that the flags are PROTOCOL_RDP, PROTOCOL_SSL and PROTOCOL_HYBRID.

for a senior

Argue from the specification's own statement that the legacy sequence cannot authenticate the server, and say which server setting turns that into a refusal.

for a principal

Frame legacy-layer retirement as a compatibility debt with an owner and a date, not as a one-off hardening step.

## The layers RDP can negotiate When an RDP client connects, the first exchange (the X.224 Connection Request and Connection Confirm) settles which **security layer** protects everything after it. The Remote Desktop Protocol specification, MS-RDPBCGR, splits the layers into two families. | Layer | Flag in the negotiation | Family | What protects the traffic | |---|---|---|---| | Standard RDP Security | `PROTOCOL_RDP` (0x00000000) | Native, section 5.3 | RDP's own RC4 encryption, or FIPS-validated methods | | TLS | `PROTOCOL_SSL` (0x00000001) | Enhanced RDP Security, section 5.4 | TLS, an external protocol | | CredSSP | `PROTOCOL_HYBRID` (0x00000002), and `PROTOCOL_HYBRID_EX` (0x00000008) | Enhanced RDP Security | TLS first, then Kerberos or NTLM inside it | The specification lists further external protocols, among them RDSTLS, which are outside this answer. The flag table names TLS 1.0, 1.1 and 1.2 for `PROTOCOL_SSL`, while the later text of section 5.4.5.1 says TLS 1.0 to 1.3. ## What Standard RDP Security does Standard RDP Security is implemented inside RDP itself. - **Cipher.** Session keys are used with RC4. The key sizes the specification names are 40, 56 and 128 bit (`ENCRYPTION_METHOD_40BIT`, `_56BIT`, `_128BIT`), plus `ENCRYPTION_METHOD_FIPS` for FIPS-validated methods. - **Keys.** The server sends its public key and a 32-byte random value in the Server Security Data; the client returns its own 32-byte random, encrypted with that public key, in the Security Exchange PDU. The two random values produce the session keys. - **Level.** The server is configured with an encryption level (`ENCRYPTION_LEVEL_NONE`, `LOW`, `CLIENT_COMPATIBLE`, `HIGH`, `FIPS`, values 0 to 4) and tells the client the method and level it selected. - **Server identity.** Section 1.3.1.2 states that the RDP connection sequence does not provide any mechanism that ensures the identity of the server is authenticated, and that it is therefore vulnerable to man-in-the-middle attacks. ## What the Enhanced layers change With Enhanced RDP Security, section 5.4 says all security operations (encryption and decryption, integrity checks, server authentication) are implemented by the external protocol. The server's security data then carries `ENCRYPTION_METHOD_NONE` and `ENCRYPTION_LEVEL_NONE` (0), and the Security Exchange PDU is not sent, because RDP no longer derives its own session keys. Section 5.4.5.1 says TLS was added to RDP to enable authentication of the remote computer's identity, mitigating man-in-the-middle attacks. CredSSP always runs the TLS handshake first; Kerberos or NTLM then authenticates the user inside the TLS channel. How that credential exchange works is a separate topic. ## Why the legacy layer is the one to retire 1. **No server authentication from the sequence itself.** A client cannot tell the server it meant to reach from something else relaying the connection. 2. **Aged cryptography.** The specification's glossary says RC4 weaknesses make it vulnerable to various cryptographic attacks, and the specification still defines the 40- and 56-bit key sizes and the Low level (which, per the product documentation, protects only client-to-server data). 3. **It is the compatibility layer.** The negotiation-based approach exists, in the specification's words, to provide backward compatibility with previous RDP implementations. Allowing it keeps old clients working and keeps the weakest option available. 4. **Microsoft's own policy page** describes native RDP encryption, as opposed to SSL encryption, as not recommended. ## How to say it in an interview 1. Name the three layers and their flags. 2. Say who does the cryptography: RDP itself for the legacy layer, an external protocol for the other two. 3. Say what the legacy layer cannot do: authenticate the server. 4. Say what you do about it: require TLS or CredSSP on the server, and treat clients that cannot offer them as a retirement project. ## A worked reading A hospital audit lists a kiosk fleet that "still negotiates Standard RDP Security". Read it in three steps. 1. The finding names the layer, so the cipher is RDP's own RC4 (or FIPS-validated methods at the FIPS level), and the server's identity is not authenticated by the sequence. 2. The level in force is a server setting, so the question to ask is whether the server allows the legacy layer at all, not which level it uses. 3. The fix is a server requirement for TLS or CredSSP; the kiosk clients that cannot meet it become the migration list. Two confusions to avoid: - **Layer against level.** The layers are three (native, TLS, CredSSP); the levels (Low, Client Compatible, High, FIPS) apply inside the native layer only. - **Encrypted against authenticated.** A legacy session is encrypted; what it lacks is a way for the client to check who the server is. ## What it leaves open Requiring an Enhanced layer does not by itself decide which TLS versions are acceptable, whether the certificate is trusted, or how the user is authenticated; those are separate settings and separate topics. It also does not fix a client that cannot offer anything but `PROTOCOL_RDP`; that client is refused, which is the intended outcome and also the migration cost.

  • If the legacy layer is the weak one, why does the specification keep it at all?
    Backward compatibility. The negotiation-based approach lets a client list the protocols it supports and a server pick one, so older RDP implementations that know only Standard RDP Security still connect. If the server's response carries no negotiation data at all, the client assumes the server does not support Enhanced RDP Security and that `PROTOCOL_RDP` is in force. The compatibility is the cost: a server that still allows it keeps the weakest option available.
  • Does choosing TLS remove the need to set an RDP encryption level?
    Yes for the level as a protection. With Enhanced RDP Security the server's security data carries encryption method and level 0 (`NONE`), and Microsoft's policy page says the client connection encryption level setting applies only to native RDP encryption and does not apply to SSL encryption. The level still matters on any listener that allows the legacy layer, which is why leaving it high is not evidence that TLS is in use.

  • Says RDP traffic is always TLS-protected by default everywhere.
  • Treats the encryption level setting as controlling TLS sessions.
  • Believes FIPS level makes Standard RDP Security equal to TLS.
  • Claims CredSSP replaces TLS rather than running a TLS handshake first.
  • Says 128-bit RC4 makes the legacy layer safe to leave on.
  • Confuses the three layers with the three encryption levels.
open as a page

What does an RD Gateway do for Remote Desktop users, and why does it keep TCP 3389 off the internet?

level: juniorimportance: must knowfreq 62%

basics

~20 s

An RD Gateway takes the user's connection over HTTPS on port 443, authenticates and authorizes it, then relays the RDP traffic to the internal host. Only the gateway faces the internet; the hosts keep 3389 private.

open as a page

In RDP, what does Network Level Authentication do before a session exists, and why does authenticating first reduce what an unauthenticated client can reach?

level: juniorimportance: must knowfreq 62%

basics

~20 s

Network Level Authentication runs CredSSP inside TLS, so the client proves its identity before the server sets up an RDP session. A client that fails authentication never reaches the MCS connect exchange or any later RDP phase.

open as a page

Why is a Windows server answering Remote Desktop on TCP 3389 from the internet treated as an incident route, and what do you fix first?

level: juniorimportance: must knowfreq 68%

basics

~20 s

An internet-reachable RDP listener lets strangers try pre-authentication flaws such as BlueKeep and guess passwords around the clock. Remove direct exposure with a gateway or VPN, require NLA and multifactor authentication, and keep the host patched.

open as a page

In RDP, which virtual channels carry clipboard, drive, printer and audio redirection, and which of them move data out of a session?

level: juniorimportance: must knowfreq 55%

basics

~20 s

A virtual channel is a named data stream inside the one RDP connection. The clipboard uses CLIPRDR and drives use RDPDR, which also carries printers; audio output uses RDPSND. Clipboard and drives move data both ways.

open as a page

In RD Gateway, what do RD CAP and RD RAP each decide, and in which order does a connection meet them?

level: middleimportance: must knowfreq 34%

basics

~20 s

A connection meets RD CAP first, which decides whether the user may connect through the gateway at all (user and device groups, authentication method). RD RAP then decides which internal target hosts that user may reach.

open as a page

What made BlueKeep (CVE-2019-0708) wormable, which Windows releases did it affect, and how do CVE-2019-1181 and CVE-2019-1182 (DejaBlue) differ?

level: middleimportance: must knowfreq 52%

basics

~20 s

BlueKeep is a pre-authentication remote code execution flaw in Remote Desktop Services, which CISA calls wormable, affecting older releases such as Windows 7 and Server 2008 R2. CVE-2019-1181 and CVE-2019-1182 are similar flaws in newer releases.

open as a page

In RDP, how does disconnecting differ from logging off, and why is a closed session still there the next morning?

level: middleimportance: must knowfreq 52%

basics

~20 s

Disconnecting ends only the network connection; the session stays on the host with its programs running. Logging off ends the session. Closing the RDP client window is a disconnect, so the session remains until sign-out or a time limit.

open as a page

A review finds RDP session hosts that still allow Standard RDP Security: why does that leave a downgrade open, and what server setting closes it?

level: seniorimportance: must knowfreq 41%

basics

~20 s

While a server allows PROTOCOL_RDP, the layer is chosen before any protection exists, so a session can end on the legacy layer. Requiring TLS or NLA on the server makes it refuse legacy-only offers with a failure code.

open as a page

Why can an administrator connected to a host over RDP with Remote Credential Guard reach a file share from that session, while one connected with Restricted Admin mode cannot?

level: seniorimportance: must knowfreq 40%

basics

~20 s

Neither mode sends a credential to the host. Restricted Admin reaches other resources as the host's own identity, so there is no single sign-on or multi-hop. Remote Credential Guard redirects Kerberos requests to the client, so the user's identity works onward.

open as a page

In the RDP X.224 negotiation, what do requestedProtocols, selectedProtocol and a failure code such as SSL_REQUIRED_BY_SERVER each tell the client?

level: middleimportance: should knowfreq 33%

basics

~20 s

The client lists the protocols it supports in requestedProtocols; the server picks one in selectedProtocol, or refuses with a failure code. A code such as SSL_REQUIRED_BY_SERVER tells the client the server's policy, and the server then closes the connection.

open as a page

In Remote Desktop Services, what do the Connection Broker and RD Web Access each do for a collection served by two session hosts?

level: middleimportance: should knowfreq 38%

basics

~20 s

The Connection Broker places each new connection on a session host in the farm, can balance the load, and returns a user whose session disconnected to the host still holding it. RD Web Access publishes the desktops and apps.

open as a page

Which transports can RD Gateway use to carry RDP, and what do TCP 443, UDP 3391 and port 3388 each do?

level: middleimportance: should knowfreq 12%

basics

~20 s

RD Gateway carries RDP over RPC over HTTP, over an HTTPS tunnel (443 by default), or over a DTLS-secured UDP channel on 3391. The RPC transport's specification also lists 80 and 3388, the gateway's own RPC listener.

open as a page

Inside NLA, in what order do TLS, SPNEGO, the public-key binding and credential delegation run in CredSSP, and what does each step settle?

level: middleimportance: should knowfreq 38%

basics

~20 s

CredSSP builds a TLS channel, runs SPNEGO inside it to authenticate with Kerberos or NTLM, has the client prove the server's TLS public key matches that authentication, and only then delegates the user's credentials to the server.

open as a page

What did the 2018 CredSSP update for CVE-2018-0886 change, and what do the Force Updated Clients, Mitigated and Vulnerable settings mean for a client and a server?

level: middleimportance: should knowfreq 27%

basics

~20 s

The March 2018 update corrected how CredSSP validates requests (CVE-2018-0886) and added the Encryption Oracle Remediation policy. Force Updated Clients blocks insecure fallback and refuses unpatched clients, Mitigated blocks fallback but accepts them, and Vulnerable allows fallback.

open as a page

How do you find RDP exposure on your own address ranges and recognise password guessing against it in Windows event 4625?

level: middleimportance: should knowfreq 40%

basics

~20 s

Compare your inventory of external address ranges with firewall and NAT rules and an external port check of those ranges. On hosts, read event 4625: logon type, failure status and source address show bursts of guessing and lockouts.

open as a page

In RDP, how do static virtual channels differ from dynamic virtual channels, and where does the limit of 31 apply?

level: middleimportance: should knowfreq 32%

basics

~20 s

Static virtual channels are named during connection setup, at most 31, and joined before the session starts. Dynamic ones are opened later inside the static channel DRDYNVC, and the 31 limit does not cover them.

open as a page

How do you bind a CA-issued certificate to an RDP listener, and what does the client's warning for the default self-signed one tell users?

level: seniorimportance: should knowfreq 30%

basics

~20 s

Import the certificate into the computer account's Personal store and point the listener at its thumbprint through the SSLCertificateSHA1Hash property or registry value. The default self-signed certificate triggers a warning because the client cannot verify who the server is.

open as a page

When Remote Desktop Connection opens a signed .rdp file after the April 2026 update, what does the dialog prove, and what does it not?

level: seniorimportance: should knowfreq 15%

basics

~20 s

Since April 2026 the client shows a dialog on every .rdp open, naming the publisher and each requested redirection, all off by default. A signature proves who signed and that the file is unaltered, not that either is safe.

open as a page

After the 2018 CredSSP update, patched Windows clients cannot open RDP sessions to servers not yet updated and show an encryption oracle error; what explains it and how do you sequence the fix?

level: seniorimportance: should knowfreq 17%

basics

~20 s

Patched clients refuse the insecure CredSSP version that unpatched servers offer and log event 6041. Patch servers first, then clients, run Mitigated, and enforce Force Updated Clients only once every client supports the newest version.

open as a page

Event 4625 bursts hit an internet-facing RDP host: what does an account lockout threshold stop, what does it cost, and what does it leave open?

level: seniorimportance: should knowfreq 30%

basics

~20 s

A lockout threshold locks an account after a set number of failed sign-ins, which limits online guessing. It costs real users and lets outsiders lock accounts on purpose, and it does not counter offline attacks or remove the exposure.

open as a page

After a network drop, an RDP client returns to its locked session with no credential prompt: what replaced the credentials, and what do events 4779 and 4778 show?

level: seniorimportance: should knowfreq 17%

basics

~20 s

After logon the server gives the client an auto-reconnect cookie bound to the session. After a network failure the client sends a modified copy instead of credentials. Events 4779 and 4778 record the disconnect and the return.

open as a page

You must retire Standard RDP Security across a fleet that includes old clients and clinical kiosks: how do you sequence it, and what do you do with devices that cannot change?

level: principalimportance: should knowfreq 15%

basics

~20 s

Inventory clients, fix certificates, then require TLS and NLA host group by host group, measuring refusals by failure code. Put devices that cannot change in a restricted host group at High, with an owner and an end date.

open as a page

A company replaces a router port-forward of 3389 with an RD Gateway; since the gateway is now the internet-facing server, what must its owner commit to?

level: principalimportance: should knowfreq 9%

basics

~10 s

Treat the gateway as the one internet-facing server: a publicly trusted certificate, prompt updates, MFA, narrow CAP and RAP policies and monitoring. The 2020 gateway flaws needed no login, so MFA cannot replace patching.

open as a page

For administrator RDP to tier-0 jump hosts that must hold no reusable credential, which delegation mode do you mandate, how do you enforce and audit it, and what does each choice cost?

level: principalimportance: should knowfreq 13%

basics

~20 s

Mandate a credential-less mode: Restricted Admin when work stays on the jump host, Remote Credential Guard when administrators need onward single sign-on over Kerberos. Enable it on the host, require it by client policy, and audit event 4624.

open as a page

A retailer's review finds one store server answering on 3389 from the internet, unpatched since 2019, among 300 stores: how do you close the exposure and prove it stays closed?

level: principalimportance: should knowfreq 22%

basics

~10 s

Inventory every external range, check the exposed server for misuse, close inbound RDP at each site, offer a gateway or VPN with multifactor authentication, require NLA and patches, and verify continuously from outside.

open as a page

How would you set RDP redirection policy for contractors, employees and administrators who share session hosts, given what each setting costs users?

level: principalimportance: should knowfreq 26%

basics

~20 s

Give each group only the redirections its work needs, set the clipboard direction by direction, and remember drive redirection is host-wide, so groups with different drive rules need different hosts. Record each kept channel as residual risk.

open as a page

In RDP, what does the 'Set client connection encryption level' policy set to High actually control, and does it cover sessions that negotiate TLS?

level: middleimportance: nice to knowfreq 19%

basics

~20 s

The policy sets the encryption level only for native RDP encryption, Standard RDP Security. It does not apply to TLS sessions, so High is no evidence of TLS. FIPS comes from the system-wide cryptography setting.

open as a page

In RDP's connection sequence, which phases fix and join the static virtual channels, and what must finish before graphics flow?

level: middleimportance: nice to knowfreq 15%

basics

~20 s

The client requests its static channels in the basic settings exchange and joins them in the channel connection. Input starts after Confirm Active and graphics after the Font List PDU, both late in the sequence.

open as a page

Who should hold the right to log on through Remote Desktop Services, and what separate control governs an administrator shadowing a user's live session?

level: seniorimportance: nice to knowfreq 14%

basics

~20 s

The Allow log on through Remote Desktop Services right decides who may open an RDP logon; restrict it to groups that need it. Shadowing, started with mstsc /shadow, is a separate administrator feature with its own consent and control rules.

open as a page