Invalid signatures

[jtallinger]

Hi,

(I move conversation over here as it is mostly just related to SigTool process now).

That last change in SigTool somehow made it worse I think. Now every file is flagged for something. var dump on $VN shows for example:

array(2) { [0]=> string(25) "� ��Java.Trojan.Boonana-6" [1]=> string(4) "2005" }

With index 1 being just "2005" sure enough it will flag almost any file. I'm sure that is not correct signature?

Not many faulty signatures now (have not checked how many is bypassed due to !isset($VN[1])):

Anti-virus: Signature errors during file scan: 2 of 35563
Last error message: preg_match(): Compilation failed: number too big in {} quantifier at offset 78
Last signature: 7265766973696f6e20383630206c747b717569747d69662f417b2f4434303c(?:..){122880,}39303930393039303930393039303930393039304538303030303030303035

[jtallinger]

I guess we should never store a signature without a valid/empty signature hash? Now it seems that may happen.

For the scanning itself, don't know how risky this change is or if this is enough to handle faulty signatures (functions.php/row 3141):

private function splitSigParts($sig, $max = -1)
{
        // PREG_SPLIT_NO_EMPTY removed
        return preg_split('~(?<!\?|\<)\:~', $sig, $max);
}

Then VN[1] will be returned but as empty value, which means VN[2] will not wrongly be used as signature hash. That along with following change (functions.php/row 2534):

// Filter out faulty signatures
if (!isset($VN[1]) || empty($VN[1])) {
    continue;
}

[jtallinger]

Added some additional logging to missing and empty signatures during scan, which brought up the total invalid signatures:

Anti-virus: Signature errors during file scan: 83 of 35561
Last error message: Signature hash missing or empty
Last signature: � ��Vbs.Trojan.PowerShell-5705194-0:

~80 empty signatures found

[Maikuolan]

@jtallinger,

I've made a few minor changes to the SigTool since the time that this issue was first created, since we'd last discussed it. If you get a spare moment, would you be able to test it out again, to let me know whether these invalid signatures still persist in signature files freshly generated by SigTool as of the current latest commit? Cheers. :-)

It appears to be okay now at my end, but I'd be more confident about the problem being actually solved if it could be confirmed by way of a second opinion (after which, I could close this off, or.. if the invalid signatures still persist, I could try to explore some new avenues of testing and so on).

[Maikuolan]

Marking as fixed and closing.