skip to content

How do Burp Suite Sequencer's character-level and bit-level analyses differ, and why might the character-level tab be empty?

level: middleimportance: nice to knowfreq 9%

answer

  1. per position against per bit
  2. sample size gates a tab
  3. FIPS is one family of tests
  4. tests need equal-length tokens
  5. decode before analysing

basics

~20 s

Character-level tests (count and transition analysis) examine each position of the raw token; bit-level tests convert tokens to bits and run FIPS, spectral, correlation and compression tests. Character-level tests are disabled at low sample sizes as unreliable.

solid answer

~40 s

In Burp Suite 2026.9, Sequencer reports two analysis levels. **Character-level** analysis looks at each character position of the raw token with two tests, character count and character transition. **Bit-level** analysis converts each token into bits and runs the FIPS monobit, poker, runs and long-runs tests, plus spectral, correlation and compression tests. The character-level tab can be empty because Sequencer disables those tests at low sample sizes: they are unreliable when the character set is large relative to the sample, for example 64 different characters at each position with only 100 tokens. Token handling settings matter too: short tokens are padded so all are the same length, and **Base64-decode before analyzing** generally improves accuracy.

go deeper

for a junior

Say there are two analysis levels, characters and bits, and that tests need a large enough sample to be reported.

for a middle

Name the two character-level tests, the family of bit-level tests, and the rule that disables character-level tests when the character set is large relative to the sample.

for a senior

Explain how the token handling settings (padding, Base64 decoding) change what is measured, and record any disabled tests next to the result.

for a principal

Keep a standard for what a Sequencer write-up must hold so that results from different assessors can be compared: sample size, settings, tests enabled and levels read.

## Two levels of analysis Burp Suite's **Sequencer** runs its tests on two levels, and the results window has a tab for each: **Character-level analysis** and **Bit-level analysis**. Each tab carries two summary charts (a significance levels chart and an effective entropy chart) and the details of every individual test, including anomalies and a description of what the test checks. This answer follows the documentation of Burp Suite Professional and Community 2026.9. | Level | What it examines | Tests | |---|---|---| | **Character-level** | Each character position of the token in its raw form | Character count analysis; character transition analysis | | **Bit-level** | Each token converted into a set of bits, tested at each bit position | FIPS monobit, FIPS poker, FIPS runs, FIPS long runs, spectral, correlation, compression | ## What each test looks at - **Character count** checks how characters are distributed at each position; a random sample is roughly uniform. - **Character transition** checks what follows a character at a position in the next token of the sample. - **FIPS monobit** counts ones and zeros at each bit position. - **FIPS poker** counts the 16 possible four-bit values and runs a chi-square calculation. - **FIPS runs and long runs** count runs of equal bits and measure the longest run. The long-runs test records a fail only when the longest run is overly long, although a run that is too short is also a sign of non-randomness. - **Spectral** treats consecutive numbers as coordinates in a multi-dimensional space and looks for clusters, which can reveal non-randomness that the other tests pass. - **Correlation** looks for relationships between bit positions, which assesses the token as a whole, since the other bit-level tests treat each position in isolation. - **Compression** tries to compress the bit sequence with ZLIB; more compression means less likely to be random. Each test begins with the hypothesis that the tokens are random, and gives the probability of what it saw under that hypothesis. ## Why the character-level tab can be empty Sequencer disables the character-level tests at low sample sizes so that unreliable results do not undermine the overall conclusion. The test of reliability is the size of the character set relative to the number of tokens. The documentation's example is that you cannot draw a reliable conclusion from 100 tokens if a token uses 64 different characters at each position. So an empty tab is not a fault; it is the tool declining to state something the sample cannot support, and the fix is a larger sample. ## Token handling that changes the answer The Sequencer settings are project settings, and three of them shape how tokens are read. 1. **Pad short tokens at start / end.** The tests can only be performed on tokens of identical length, so variable-length tokens are padded. Padding at the start is usually right. 2. **Pad.** The padding character; for numeric or ASCII hex tokens, `0` is usually most appropriate. 3. **Base64-decode before analyzing.** Decoding Base64 tokens first generally improves accuracy, because it analyses what the generator produced rather than the alphabet that carried it. You can also enable or disable each test individually, change the settings on the **Analysis settings** tab and choose **Redo analysis**. That is useful for isolating the effect of an unusual characteristic. It is not a way of making a result look better: a record of which tests were switched off belongs in the write-up. ## Where the output stops being trustworthy - If a position has a character set whose size is not a round power of two, Sequencer translates the data to the nearest smaller round power of two, preserving randomness characteristics without adding bias, but the documentation notes that this is likely to introduce some inaccuracies. - The FIPS results are adjusted to work with the sample sizes you provide; results compliant with the formal FIPS standard need 20,000 tokens. - Each bit-level test works on individual bit positions in isolation, so a sample with the same bit value at every position can appear to contain more entropy than shorter tokens with varied values; the correlation test is what assesses the token as a whole. ## Common slips - Expecting the character-level tab to appear whatever the sample size. - Treating FIPS pass or fail as formal compliance on a sample smaller than 20,000. - Switching off failing tests until the summary improves, without recording it. - Forgetting to decode Base64 tokens before analysis.

  • Why does Sequencer pad short tokens before analysis, and where is padding usually applied?
    The statistical tests can only run on tokens of identical length, so variable-length tokens are padded with a chosen character. Padding at the start is usually right, and 0 usually suits numeric or ASCII hex tokens. A padded count is reported in the sample summary, so a high number there tells the reviewer to read the result cautiously.
  • Why would you disable individual tests and redo an analysis, and what is the caution?
    After an initial run with all tests enabled, you can switch tests off to reflect what you know about the tokens or to isolate the effect of an unusual characteristic, then use Redo analysis. The caution is to record which tests were off and why, since disabling tests until the result improves would misstate the evidence.

  • Expects the character-level tab to appear whatever the sample size
  • Says Sequencer pads tokens only for display, not for the tests
  • Treats FIPS pass or fail as formal compliance on any sample size
  • Disables failing tests until the overall result improves, without noting it
  • Analyses Base64 tokens as raw text without considering the decode setting

Assumes Burp Suite Professional / Community 2026.9 documentation for Sequencer's tests and settings pages; the test list and the low-sample rule are as that release's pages state them.