Table 2 Procedure of EMBF algorithm.

Let B1 = ⌊n/L⌋, B2 = ⌊m/L⌋, E = MAX(M, G);

1. For offset = 1 to L do

   1.1 Divide S [offset, n] into L bps blocks, as S_offset = {s_offset,1,..., s_{offset, B1}};

   1.2 Convert S_offsetto frequency vectors, as ES_offset = {es_offset,1,..., es_{offset, B1}};

2. For offset = 1 to L do

   2.1 Sequentially choose B2 blocks from ES_offset, and set the start position as p; Using all possible combinations to get B2-E blocks. And combine them as ADDR variable. Set the remaining E blocks as r;

   2.2 Mapping pair (ADDR,(r, p)) into a hash map M, and chaining possible conflicts;

   2.3 Iteratively scan ES_offset for next B2 blocks in ES_offset;

3 First level filtering process

   3.1 Divide T into L bps blocks, as T = {t₁, t₂,..., t_B2} and convert them to frequency vector as ET = {et₁, et₂,..., et_B2};

   3.2 Choose B2-E blocks from ET and combining them as ADDR variable, set the remaining E blocks as t;

   3.3 Query ADDR in M and pass all returned results as R = {(r₁, p₁), (r₂, p₂),...... } to step 4;

4 For i = 1 to | R| do

   if ECD(t, r_i) < E then record p_i;

Output: All recorded p _i from step 4.