I wonder is this might be a good topic for the next infra-wg. One thing we
may be able to do is fix up the regex so stuff such as 'yum install curl'
or 'apt-get install wget' don't cause false alarms.

The good thing with the framework is the regexs are easy to get at, so its
easy for anyone to push a patch with a better fine tuned regex.


On Thu, Mar 8, 2018 at 3:07 PM, SULLIVAN, BRYAN L (BRYAN L) <
[email protected]> wrote:

> Fatih,
>
>
>
> I think the problem is that it’s very difficult to differentiate between
> use of wget/curl etc for:
>
>    - Pulling in components/data from external sources, as part of a
>    deploy or test process, or even by design of the component being developed
>    (e.g. a config file is pulled from some repo/place and used in component
>    configuration)
>    - Use of APIs exposed by the NFV platform or applications running
>    under it
>
>
>
> Requiring all the latter uses to be allowed through regexp rules would be
> the onerous part.
>
>
>
> Thanks,
>
> Bryan Sullivan | AT&T
>
>
>
> *From:* Fatih Degirmenci [mailto:[email protected]]
> *Sent:* Thursday, March 08, 2018 7:01 AM
> *To:* SULLIVAN, BRYAN L (BRYAN L) <[email protected]>; Luke
> Hinds <[email protected]>; opnfv-tech-discuss <opnfv-tech-discuss@lists.
> opnfv.org>
>
> *Subject:* Re: [opnfv-tech-discuss] [releng][security][infra] Anteater
> Improvements
>
>
>
> Hi Brian,
>
>
>
> My comment wasn’t about the tools themselves but what they are used for
> and to be honest the suggestion is nothing sort of heavy-handed approach.
>
>
>
> If someone includes something in an artifact that is consumed by someone
> else for different purposes, we have responsibility to them that we will
> always be able to recreate that artifact for different purposes such as
> troubleshooting.
>
> If people hook their own machines to store artifacts that go into final
> OPNFV release and if these machines disappear, we will have no possibility
> to recreate what we released.
>
> Apart from that, if people want to rebuild something for some reason, they
> will be unable to do that since the dependencies will not be available
> anymore.
>
> These concerns are based on what we found while looking at our earlier
> releases so I’m not talking about something that’s possible to happen but a
> real issue that happened and we had to intervene.
>
>
>
> If projects guarantee the long term availability of the artifacts on the
> original location (ie until the EOL of that specific release) and they ask
> for exception, there will be no blocking.
>
>
>
> As you said, you are free to the promote tools and approaches like how I
> am doing here since I believe they are important to be shared.
>
>
>
> /Fatih
>
>
>
> *From: *"SULLIVAN, BRYAN L (BRYAN L)" <[email protected]>
> *Date: *Thursday, 8 March 2018 at 15:31
> *To: *Fatih Degirmenci <[email protected]>, Luke Hinds <
> [email protected]>, "[email protected]" <
> [email protected]>
> *Subject: *RE: [opnfv-tech-discuss] [releng][security][infra] Anteater
> Improvements
>
>
>
> I do recommend that we rely upon tools that can focus on the trust of
> specific sources, and not the use of platform capabilities such as curl,
> wget, etc. These (curl, wget, etc) are tools that can be used for many
> purposes inside an application like an OPNFV platform, or its
> deployment/testing tools. The flagging/blocking of patches just because the
> code contains use of wget/curl/etc is an onerous and heavy-handed approach
> to the goal. Having to create regexp rules for each valid use of wget/curl
> is a non-starter for me and likely many others in this project. Thus I
> support Luke’s plan, and intend to promote these same tools and approaches
> for use in the Acumos and other LFN projects.
>
>
>
> Thanks,
>
> Bryan Sullivan | AT&T
>
>
>
> *From:* [email protected] [
> mailto:[email protected]
> <[email protected]>] *On Behalf Of *Fatih
> Degirmenci
> *Sent:* Thursday, March 08, 2018 6:12 AM
> *To:* Luke Hinds <[email protected]>; opnfv-tech-discuss <
> [email protected]>
> *Subject:* Re: [opnfv-tech-discuss] [releng][security][infra] Anteater
> Improvements
>
>
>
> Hi Luke,
>
>
>
> I have few comments and followup questions regarding this:
>
> “This in turn means we won't raise alarms over curl, git clone and wget
> and will instead check the IP addresses or URLS that those commands query.
> This should make anteater a lot less chatty at gate.”
>
>
>
> You might remember that one of the reasons we have checks for curl/wget is
> to find out if projects pull artifacts from unknown IPs during
> build/deployment/testing.
>
> These are not malicious but we have seen that few of the IPs where the
> projects fetch the artifacts belong to non-production/personal devices that
> tend to disappear over time.
>
> As you know, this is an important issue from reproducibility and
> traceability perspectives.
>
>
>
> Now the questions are;
>
> Assuming the IPs are not explicitly added to exception list for the
> corresponding project, do you mean that we will stop flagging changes/files
> that contain wget/curl against unknown IPs if they are not marked as
> malicious on VirusTotal?
>
> We also had plans to make anteater checks voting/blocking. Will we discard
> this plan since wget/curl against IPs are not even planned to be flagged?
>
>
>
> /Fatih
>
>
>
> *From: *<[email protected]> on behalf of Luke
> Hinds <[email protected]>
> *Date: *Thursday, 8 March 2018 at 14:02
> *To: *"[email protected]" <opnfv-tech-discuss@lists.
> opnfv.org>
> *Subject: *[opnfv-tech-discuss] [releng][security][infra] Anteater
> Improvements
>
>
>
> Hello,
>
> I have some changes to improve the reporting ability and hopefully tone
> down the false positives.
>
> Aneater will now interface with the VirusTotal public API:
>
> 1. If anteater finds a public IP address, the DNS history will be quiered
> to see if the IP has past or present associations with malicious domains.
>
>
>
> 2. If a URL is found, it is checked against the VirusTotal API to see if
> its marked as malicous.
>
> 3. Binaries will be sent to VirusTotal for a scan by the aggregation of
> scanners hosted there.
>
> For anyone wanting a demo, please see the following:
>
> https://asciinema.org/a/JfzUPWpBGm0wDKPCN3KlK2DK0
> <https://urldefense.proofpoint.com/v2/url?u=https-3A__asciinema.org_a_JfzUPWpBGm0wDKPCN3KlK2DK0&d=DwMGaQ&c=LFYZ-o9_HUMeMTSQicvjIg&r=ML-JPRZQOfToJjMwlJLPlcWimAEwMA5DZGNIrk-cgy0&m=Phto5YquFG94SB6m2dgJ83G2mmGj0cNr14cgRF9LHwA&s=2OjBdIMqqEr-gCgFDNZLOLPPdiF3zOemzOq73XIqiSU&e=>
>
> I will work with various people to get this rigged into CI.
>
> This in turn means we won't raise alarms over curl, git clone and wget and
> will instead check the IP addresses or URLS that those commands query. This
> should make anteater a lot less chatty at gate.
>
> Cheers,
>
> Luke
>



-- 
Luke Hinds | NFV Partner Engineering | CTO Office | Red Hat
e: [email protected] | irc: lhinds @freenode | t: +44 12 52 36 2483
_______________________________________________
opnfv-tech-discuss mailing list
[email protected]
https://lists.opnfv.org/mailman/listinfo/opnfv-tech-discuss

Reply via email to