如何基于ssdeep进行群集？

我已经生成了ssdeep文件并保存在csv文件中。

我在perl脚本解析文件，如下所示：

foreach(@all_lines) 
{ 
    chomp; 
    my $line = $_; 

    my @split_array = split(/,/, $line); 
    my $md5 = $split_array[1]; 
    my $ssdeep = $split_array[4]; 

    my $blk_size = (split(/:/, $ssdeep))[0]; 

    if($blk_size ne "") 
    { 
     my $cluster_id = check_In_Cluster($ssdeep); 

     print WFp "$cluster_id,$md5,$ssdeep\n"; 
    } 
}

这也检查ssdeep是否存在于先前群集组，如果不是创造了新的团体。

代码chec_In_Cluster

my $ssdeep = shift; 
my $cmp_result; 

if($cluster_cnt > 0) { 
    $cmp_result = ssdeep_compare($MRU_ssdeep, $ssdeep); 
    if($cmp_result > 85) { 
     return $MRU_cnt; 
    } 
} 

my $d = int($cluster_cnt/4); 
my $thr1 = threads->create(\&check, 0, $d, $ssdeep); 
my $thr2 = threads->create(\&check, $d, 2*$d, $ssdeep); 
my $thr3 = threads->create(\&check, 2*$d, 3*$d, $ssdeep); 
my $thr4 = threads->create(\&check, 3*$d, $cluster_cnt, $ssdeep); 

my ($ret1, $ret2, $ret3, $ret4); 
$ret1 = $thr1->join(); 
$ret2 = $thr2->join(); 
$ret3 = $thr3->join(); 
$ret4 = $thr4->join(); 
if($ret1 != -1) { 
    $MRU_ssdeep = $ssdeep; 
    $MRU_cnt = $ret1; 
    return $MRU_cnt;  
} elsif($ret2 != -1) { 
    $MRU_ssdeep = $ssdeep; 
    $MRU_cnt = $ret2; 
    return $MRU_cnt;  
} elsif($ret3 != -1) { 
    $MRU_ssdeep = $ssdeep; 
    $MRU_cnt = $ret3; 
    return $MRU_cnt;  
} elsif($ret4 != -1) { 
    $MRU_ssdeep = $ssdeep; 
    $MRU_cnt = $ret4; 
    return $MRU_cnt;  
} else { 
    $cluster_base[$cluster_cnt] = $ssdeep; 
    $MRU_ssdeep = $ssdeep; 
    $MRU_cnt = $cluster_cnt; 
    $cluster_cnt++; 
    return $MRU_cnt;  
}

和再检查一下代码：

sub check($$$) { 
    my $from = shift; 
    my $to = shift; 
    my $ssdeep = shift; 

    for(my $icnt = $from; $icnt < $to; $icnt++) { 
     my $cmp_result = ssdeep_compare($cluster_base[$icnt], $ssdeep); 
     if($cmp_result > 85) { 
      return $icnt; 
     } 
    } 
return -1; 
}

但这个过程需要非常多的时间（因为要用8-9Hours 20-30MB csv文件）。我也尝试使用多线程，而在群集中检查，但没有太多的帮助，我从这。

由于他们不需要csv解析器，如Text :: CSV（因为对csv的操作较少），我没有使用它。

任何人都可以解决我的问题吗？是否有可能使用hadoop或其他框架进行基于ssdeep的分组？

来源

2014-02-14 Sanjay Bhosale