Checksum

Adler-32

Designed by Mark Adler for zlib (1995). Adler-32 is a Fletcher-style checksum: two 16-bit running sums combined into a 32-bit output. It is faster than CRC-32 by a wide margin but noticeably weaker for short inputs , the trade-off is deliberate for the inner loop of DEFLATE.

How it works, in one paragraph

Maintain two 16-bit accumulators, a and b, both modulo 65521 (the largest prime < 2¹⁶). For each byte x: a = (a + x) mod 65521, b = (b + a) mod 65521. Output (b << 16) | a. The trailing b integrates position information so byte reordering changes the digest.

At a glance

Where it shows up

• zlib / DEFLATE , every zlib stream ends with an Adler-32 of the uncompressed data.
• PNG , uses zlib internally, so PNG files inherit Adler-32 (alongside their own CRC-32 chunks).
• Rsync , rolling Adler variant as the fast pass in the delta-transfer algorithm.

Limitations

• For inputs under a few hundred bytes, Adler-32’s effective range is smaller than CRC-32’s, increasing false-positive integrity checks.
• Modulo 65521 means the running sums can’t use the full 32-bit space, slightly reducing distribution.
• Trivially collidable by an attacker; never use for adversarial integrity.

References

• RFC 1950 , zlib compressed data format (defines Adler-32)
• Wikipedia , Adler-32
• Fletcher · CRC-32