skip to content

WPA

TKIP as the emergency retrofit over WEP's RC4 engine: per-packet key mixing, the Michael MIC with countermeasures, the TSC replay counter, and why it was retired. A design lesson in compatibility.

on this pageshow

explore

questions

6

What was TKIP in the first WPA generation, and why did it keep WEP's RC4 cipher while changing keys, integrity and replay checks?

level: juniorimportance: must knowfreq 62%

answer

  1. firmware update, not new hardware
  2. RC4 kept, frame processing around it changed
  3. key mixing, Michael, sequence counter
  4. interim fix, CCMP is the real answer

basics

~20 s

TKIP is the Temporal Key Integrity Protocol, an emergency retrofit that let WEP-era hardware run WPA after a firmware or driver update. It kept the RC4 engine and added per-packet key mixing, the Michael integrity check and a sequence counter.

solid answer

~40 s

TKIP was designed so access points and adapters built for WEP could be upgraded in software, because CCMP's AES needs computing resources that pre-RSN hardware cannot be assumed to have. So TKIP keeps RC4 and the WEP frame processing, and wraps three repairs around it: a fresh RC4 key for every frame, mixed from the temporal key, the transmitter address and a sequence counter; the 8-octet Michael MIC, because WEP's CRC-32 is not integrity; and the TKIP sequence counter (TSC), which makes the receiver drop replayed frames. CCMP, built on AES, replaced it as the mandatory cipher of WPA2. TKIP is a deprecated stopgap.

go deeper

for a junior

Recall that TKIP was a firmware-level retrofit for WEP hardware that kept RC4, added three repairs and was replaced by CCMP.

for a middle

Explain what each repair fixes in WEP: per-frame key mixing against the IV-prefix weakness, Michael against CRC-32 forgery, the TSC against replay.

for a senior

Frame TKIP as the design lesson in compatibility: speak to what the old hardware budget forced into the standard and why that still shows in the cipher.

for a principal

Treat any inherited compatibility mode as debt: ask what still depends on it, who owns the replacement cost and what the stopgap quietly caps.

## Why TKIP exists WEP, the original 802.11 security, was shown to be breakable in 2001, and the flaws sit in its design, not in a setting. Yet the deployed access points and adapters had a radio chip with an RC4 engine, and AES could not be assumed to fit its processing budget. IEEE 802.11i therefore defined two protocols. NIST's guide to 802.11i says TKIP "was created to allow already-deployed devices to address the numerous inadequacies of WEP" and may be implemented through software updates, while CCMP "requires computing resources that cannot be assumed on pre-RSN hardware". In nearly all cases, NIST adds, deploying CCMP means replacing some equipment. WPA, which Beck and Tews date to 2003, carried that retrofit to the market. ## What TKIP kept and what it added TKIP is WEP plus three repairs, all of which had to fit the old hardware's processing budget: - **Kept:** the RC4 stream cipher and the WEP encapsulation step, including the 4-octet ICV (a CRC-32). - **Added, per-frame key mixing:** the RC4 key is no longer the shared key with an IV stuck on the front. A mixing function builds a new key for every frame from the temporal key, the transmitter address and the TSC. - **Added, the Michael MIC:** an 8-octet keyed check over the data, the source and destination addresses and the priority, so a forger can no longer patch a CRC-32 and pass as genuine. - **Added, the TSC:** a 48-bit counter that the receiver requires to increase, so a replayed frame is dropped. | Property | WEP | TKIP | CCMP | |---|---|---|---| | Core algorithm | RC4 | RC4 | AES | | Per-packet key | Shared key joined with a 24-bit IV | TKIP mixing function | Not needed; the temporal key is used | | Integrity | Enciphered CRC-32 | Michael MIC with countermeasures | CCM | | Replay detection | None | Enforced sequence counter | Enforced sequence counter | ## What it looks like in a frame The Linux 802.11 headers give the sizes: `IEEE80211_TKIP_IV_LEN` is 8 octets (the TSC occupies 6 of them), `MICHAEL_MIC_LEN` is 8 and `IEEE80211_TKIP_ICV_LEN` is 4. The key material `WLAN_KEY_LEN_TKIP` is 32 octets: a 128-bit encryption key and two 64-bit Michael keys, one per direction. The cipher suite selector in the RSN element is 00-0F-AC:2. ## What it could not fix Because the engine stayed, so did its limits. NIST states that RC4 and Michael "both have known weaknesses" and that TKIP is not suitable for high assurance environments. RC4 does not meet FIPS requirements, so TKIP cannot be used where FIPS-validated algorithms are mandated. Michael was deliberately cheap, which is why TKIP needs a countermeasure that can itself cause an outage. The published results against TKIP are separate topics, but the pattern is the point: a compatibility constraint became a permanent design weakness. ## The design lesson Each of TKIP's weaknesses traces to one of three constraints the retrofit accepted: | Constraint | Consequence | |---|---| | Keep the RC4 engine | RC4 stays, and with it the FIPS problem | | Keep the WEP frame handling | The ICV stays, and a TKIP frame carries 20 octets of IV field, MIC and ICV against WEP's 8 | | Fit the old chips' processing budget | Michael is weak by design and needs a countermeasure that can cost availability | That is the lesson of the retrofit: compatibility bought deployability and time, and paid for it with ceilings that stayed after the cipher stopped being safe to keep. ## What the operator does 1. Treat TKIP as retired. CCMP is mandatory for any device claiming RSNA compliance, while TKIP is optional. 2. Remove it where it is still offered. NIST's guide says to disable WEP and TKIP in each access point's configuration, because otherwise stations may negotiate TKIP instead of CCMP. 3. Remember that the reference implementation agrees. The comment in hostapd's default build configuration calls TKIP "an old cryptographic data confidentiality algorithm that is not considered secure", and `CONFIG_NO_TKIP=y` (added in hostapd 2.10) removes it at build time. 4. Take the retirement timeline from the comparison of Wi-Fi security generations, not from this note.

  • If TKIP added Michael, why does a TKIP frame still carry the old 4-octet CRC-32 ICV?
    Because the retrofit runs on the unchanged WEP encapsulation: TKIP computes the Michael MIC first, then hands the data and MIC to the old WEP step, which appends the ICV and encrypts with RC4. The ICV stays as an error check on the WEP step, while integrity against forgery is now the job of Michael.
  • Why does an access point menu list 'WPA' and 'TKIP' almost as synonyms?
    In the first WPA generation TKIP was the cipher, so a label like WPA/TKIP names that whole generation. TKIP stayed optional in WPA2, though, so a WPA2 network can still be configured with it. In hostapd, for example, the WPA2 cipher setting takes its default from the first-generation one.

  • Says TKIP swapped RC4 for AES; AES arrived with CCMP.
  • Treats TKIP as WEP with a longer key.
  • Claims TKIP needed new radios, so it never ran on WEP-era gear.
  • Calls TKIP acceptable long term because it is not WEP.
  • Believes the Michael MIC replaced all of WEP's frame handling, including the ICV.
open as a page

Why is TKIP's 64-bit Michael MIC weak by design, and how does the two-failures-in-a-minute countermeasure bound that weakness?

level: middleimportance: must knowfreq 38%

basics

~20 s

Michael is a cheap keyed check sized for WEP-era hardware, so it is weak. TKIP bounds it: after two MIC failures within a minute, the access point deauthenticates its TKIP clients and stops TKIP traffic for one minute.

open as a page

In TKIP, how is each frame's RC4 key built, and how does the TKIP sequence counter let a receiver reject replayed frames?

level: middleimportance: should knowfreq 28%

basics

~20 s

Each TKIP frame gets its own RC4 key, mixed from the temporal key, transmitter address and a 48-bit sequence counter (TSC). The receiver silently drops a frame whose TSC is not higher than the last accepted on that priority channel.

open as a page

In hostapd, how can TKIP stay enabled through wpa_pairwise, rsn_pairwise or the group cipher, and what removes it for good?

level: seniorimportance: should knowfreq 12%

basics

~20 s

In hostapd, wpa_pairwise defaults to TKIP and rsn_pairwise copies it, so a WPA2 section left untouched offers TKIP, and TKIP in any enabled list becomes the group cipher too. Set rsn_pairwise=CCMP, or build with CONFIG_NO_TKIP=y.

open as a page

A hospital's WPA/TKIP SSID for bedside monitors logs two Michael MIC failures and every client then drops for a minute; what happened and what do you do?

level: seniorimportance: should knowfreq 22%

basics

~20 s

The TKIP countermeasure fired: two Michael MIC failures within a minute made the access point deauthenticate its clients and refuse TKIP traffic for 60 seconds. The log cannot separate a fault from hostile activity; investigate, then move the monitors off TKIP.

open as a page

What did the Beck-Tews result against TKIP, published in 2008, show, and what bounds how far it goes?

level: seniorimportance: nice to knowfreq 8%

basics

~20 s

Published in 2008, the Beck-Tews result gives a limited outcome against TKIP: a Michael key for one direction and one frame's keystream, enough to forge a few short frames. The countermeasure caps its decryption rate; it is not a total break.

open as a page