ssh
免密登录在带来方便的同时也带来一些问题,那就是不知道什么时间什么人利用ssh
免密通道登录服务器了,为此我们需要在sshd
的配置文件里设置好详细日志,以便日后回溯。
在CentOS
里,sshd
的日志文件是保存在/var/log/secure
里的,如果不进行特殊设置的话,它只记录一些最简单的信息,比如什么时间哪个账号被人用免密登录的方式登录了,如果这个账号的authorized_keys
里有很多key
的话,这样的log
没有办法告诉你客户到底是用哪个key
登录的。
所以,我们需要在sshd
的配置文件/etc/ssh/sshd_config
文件里找到以下项LogLevel
并把它改成LogLevel VERBOSE
,这一项的缺省值是INFO
,不够详细,所以我们需要把它改成啰嗦模式,这样可以记录更多信息。
改成VERBOSE
并重启sshd
(service sshd restart
)后,我们会在日志文件里看到类似于这样的记录:
Apr 1 10:37:06 hostname sshd[5903]: Found matching RSA key: 83:67:b5:c7:bb:17:4d:06:ca:dc:8b:ca:85:cc:0c:b1
但这样的信息明显不同于我们在authorized_keys
里存放的信息,那该怎么办呢?实际上,在sshd
的日志文件里存储的只是我们authorized_keys
的指纹信息fingerprint
,不是真正的key
,必须从authorized_keys
反算出fingerprint
来,才能做对比:
ssh-keygen -E md5 -lf /home/someuser/.ssh/authorized_keys
但是这样依然很麻烦,有没有办法直接告诉我日志里到底是谁登录的呢?为此我们还需要引入一个用 Perl 写的小程序。(原作者写的略有问题,在新版的CentOS
里必须要求附加md5
参数,为此我做了一些小的修改):
#!/usr/bin/perl -w
use strict;
use diagnostics;
use File::Temp;
# Matches Fingerprints from sshd logs (sshd on loglevel VERBOSE) against
# authorized_keys for the respective user.
die "Please specify input file!\n" unless ($ARGV[0]);
my $fingerprints;
my $nav = File::Navigate->new($ARGV[0]);
# Store publickey login events
my @lines = @{$nav->find(qr/sshd\[\d+\]: Accepted publickey for .+ from .+ port \d+/)};
# Process all publickey login events
foreach(@lines){
$nav->cursor($_);
my $line = $nav->get();
$line =~ /^(.{15}).+sshd\[(\d+)\]: Accepted publickey for (.+) from (.+) port (\d+)/;
my $date = $1;
my $pid = $2;
my $user = $3;
my $ip = $4;
my $port = $5;
my $fp = "unknown"; # (yet)
# Seek backwards to find matching fingerprint line
my $sought = 0;
while ((my $seekline = $nav->getprev()) and ($sought++ < 1000)){
if ($seekline =~ /sshd\[$pid\]: Found matching .+ key: (.+)/){
$fp = $1;
last;
}elsif($line =~ /sshd\[$pid\]: Connection from $ip port $port/){
last;
}
}
my $key = get_key($fp, $user);
print "\"$date\";\"$user\";\"$fp\";\"$key\"\n";
}
sub get_key{
my $fp = shift;
$fp = "MD5:" . $fp;
my $user = shift;
# See if FP is cached
if ($fingerprints->{$user}){
if ($fingerprints->{$user}->{$fp}){
return $fingerprints->{$user}->{$fp};
}else{
return "No matching key found.";
}
}
# Else, generate fingerprints from users authorized_keys
print STDERR "------> Reading keys for user $user\n";
my $home = (getpwnam($user))[7];
open my $fh_in, "<$home/.ssh/authorized_keys" or warn "No such file: $home/.ssh/authorized_keys\n";
while (<$fh_in>){
chomp;
next unless (/^ssh-/);
my $out_fh = File::Temp->new();
print $out_fh "$_\n";
close $out_fh;
my $fp_raw = `ssh-keygen -E md5 -lf $out_fh`;
# Second field of output has the fingerpring
my $fp = (split /\s+/, $fp_raw)[1];
$fingerprints->{$user}->{$fp} = $_;
}
if ($fingerprints->{$user}->{$fp}){
return $fingerprints->{$user}->{$fp};
}else{
return "No matching key found.";
}
}
package File::Navigate;
use strict;
use warnings;
=head1 NAME
File::Navigate - Navigate freely inside a text file
=head1 DESCRIPTION
The module is a glorified wrapper for tell() and seek().
It aims to simplify the creation of logfile analysis tools by
providing a facility to jump around freely inside the contents
of large files without creating the need to slurp excessive
amounts of data.
=head1 SYNOPSIS
use File::Navigate;
my $nav = File::Navigate->new('/var/log/messages');
# Read what's below the "cursor":
my $first = $nav->get;
# Advance the cursor before reading:
my $second = $nav->getnext;
my $third = $nav->getnext;
# Advance the cursor by hand:
$nav->next;
my $fourth = $nav->get;
# Position the cursor onto an arbitrary line:
$nav->cursor(10);
my $tenth = $nav->get;
# Reverse the cursor one line backward:
$nav->prev;
my $ninth = $nav->get;
# Reverse the cursor before reading:
my $eigth = $nav->getprev;
# Read an arbitrary line:
my $sixth = $nav->get(6);
=cut
our @ISA = qw(Exporter);
our @EXPORT_OK = qw();
our $VERSION = '1.0';
=head1 CLASS METHODS
=head2 I<new()>
Open the file and create an index of the lines inside of it.
my $mapper = File::Navigate->new($filename);
=cut
sub new($){
my $class = shift;
my $file;
unless ($file = shift){
die "No file specified\n";
}
unless (-e $file){
die "File not found: $file\n";
}
unless (-r $file){
die "File not readable: $file\n";
}
my $self = {};
$self->{'cursor'} = 1;
$self->{'lineindex'} = {};
$self->{'lineindex'}->{1} = 0;
open my $fh, "$file"
or die "Can't open $file: $!\n";
while (<$fh>){
my $thisline = $.;
my $nextline = $thisline + 1;
$self->{'lineindex'}->{$nextline} = tell $fh;
}
$self->{'length'} = scalar(keys %{$self->{'lineindex'}}) - 1 ;
$self->{'fh'} = $fh;
bless $self;
}
=head1 OBJECT METHODS
=head2 I<count()>
Returns the number of lines in the file ("wc -l")
my $lines = $nav->count;
=cut
sub length(){
my $self = shift;
return $self->{'length'};
}
=head2 I<cursor()>
Returns the current cursor position and/or sets the cursor.
my $cursor = $nav->cursor(); # Query cursor position.
my $cursor = $nav->cursor(10); # Set cursor to line 10
=cut
sub cursor($){
my $self = shift;
if (my $goto = shift){
$self->{'cursor'} = $goto;
}
return $self->{'cursor'};
}
=head2 I<get()>
Gets the line at the cursor position or at the given position.
my $line = $nav->get(); # Get line at cursor
my $line = $nav->get(10); # Get line 10
=cut
sub get($){
my $self = shift;
my $fh = $self->{'fh'};
my $getline;
$getline = $self->{'cursor'} unless ($getline = shift);
if ($getline < 1){
warn "WARNING: Seek before first line.";
return undef;
}elsif($getline > $self->{'length'}){
warn "WARNING: Seek beyond last line.";
return undef;
}
seek ($fh, $self->{'lineindex'}->{$getline}, 0);
my $gotline = <$fh>;
chomp $gotline;
return $gotline;
}
=head2 I<next()>
Advance the cursor position by one line. Returns the new cursor position.
Returns I<undef> if the cursor is already on the last line.
my $newcursor = $nav->next();
=cut
sub next(){
my $self = shift;
if ($self->{'cursor'} == $self->{'length'}){
return undef;
}
$self->{'cursor'}++;
return $self->{'cursor'};
}
=head2 I<prev()>
Reverse the cursor position by one line. Returns the new cursor position.
Returns I<undef> if the cursor is already on line 1.
my $newcursor = $nav->prev();
=cut
sub prev(){
my $self = shift;
if ($self->{'cursor'} == 1){
return undef;
}
$self->{'cursor'}--;
return $self->{'cursor'};
}
=head2 I<getnext()>
Advance to the next line and return it.
Returns I<undef> if the cursor is already on the last line.
my $newcursor = $nav->getnext();
=cut
sub getnext(){
my $self = shift;
$self->next or return undef;
return $self->get;
}
=head2 I<getprev()>
Reverse to the previous line and return it:
Returns I<undef> if the cursor is already on line 1.
my $newcursor = $nav->getprev();
=cut
sub getprev(){
my $self = shift;
$self->prev or return undef;
return $self->get;
}
=head2 I<find()>
Find lines containing given regex. Returns array with line numbers.
my @lines = @{$nav->find(qr/foo/)};
=cut
sub find($){
my $self = shift;
my $regex = shift;
my @results;
for (my $lineno = 1; $lineno <= $self->{'length'}; $lineno++){
my $line = $self->get($lineno);
if ($line =~ $regex){
push @results, $lineno;
}
}
return \@results;
}
sub DESTROY(){
my $self = shift;
close $self->{'fh'};
}
=head1 EXAMPLE
I<tac>, the opposite of I<cat>, in Perl using File::Navigate:
#!/usr/bin/perl -w
use strict;
use File::Navigate;
foreach my $file (reverse(@ARGV)){
my $nav = File::Navigate->new($file);
# Force cursor beyond last line
$nav->cursor($nav->length()+1);
print $nav->get()."\n" while $nav->prev();
}
=head1 BUGS
Seems to lack proper error handling.
=head1 LIMITATIONS
Works only on plain text files. Sockets, STDIO etc. are not supported.
=head1 PREREQUISITES
Tested on Perl 5.6.1.
=head1 STATUS
Mostly harmless.
=head1 AUTHOR
Martin Schmitt <mas at scsy dot de>
=cut
1;
为此我们给它起名叫match-ssh-keys
,赋予它可执行权限(chmod +x match-ssh-keys
),然后把它搬到/usr/local/bin
里(mv match-ssh-keys /usr/local/bin/
),这样我们以后再想查谁通过sshd
免密登录过服务器就方便了,我们只需要执行:
match-ssh-keys /var/log/secure
就可以了。