Here is my quick wrap-up of the FIRST Technical Colloquium hosted by Cisco in Amsterdam. This is my first participation to a FIRST event. FIRST is an organization helping in incident response as stated on their website:
FIRST is a premier organization and recognized global leader in incident response. Membership in FIRST enables incident response teams to more effectively respond to security incidents by providing access to best practices, tools, and trusted communication with member teams.
The event was organized at Cisco office. Monday was dedicated to a training about incident response and the two next days were dedicated to presentations. All of them focussing on the defence side (“blue team”). Here are a few notes about interesting stuff that I learned.
The first day started with two guys from Facebook: Eric Water @ Matt Moren. They presented the solution developed internally at Facebook to solve the problem of capturing network traffic: “PCAP don’t scale”. In fact, with their solution, it scales! To investigate incidents, PCAPs are often the gold mine. They contain many IOC’s but they also introduce challenges: the disk space, the retention policy, the growing network throughput. When vendors’ solutions don’t fit, it’s time to built your own solution. Ok, only big organizations like Facebook have resources to do this but it’s quite fun. The solution they developed can be seen as a service: “PCAP as a Service”. They started by building the right hardware for sensors and added a cool software layer on top of it. Once collected, interesting PCAPs are analyzed using the Cloudshark service. They explained how they reached top performances by mixing NFS and their GlusterFS solution. Really a cool solution if you have multi-gigabits networks to tap!
The next presentation focused on “internal network monitoring and anomaly detection through host clustering” by Thomas Atterna from TNO. The idea behind this talk was to explain how to monitor also internal traffic. Indeed, in many cases, organizations still focus on the perimeter but internal traffic is also important. We can detect proxies, rogue servers, C2, people trying to pivot, etc. The talk explained how to build clusters of hosts. A cluster of hosts is a group of devices that have the same behaviour like mail servers, database servers, … Then to determine “normal” behaviour per cluster and observe when individual hosts deviate. Clusters are based on the behaviour (the amount of traffic, the number of flows, protocols, …). The model is useful when your network is quite close and stable but much more difficult to implement in an “open” environment (like universities networks).
Then Davide Carnali made a nice review of the Nigerian cybercrime landscape. He explained in details how they prepare their attacks, how they steal credentials, how they deploy the attacking platform (RDP, RAT, VPN, etc). The second part was a step-by-step explanation how they abuse companies to steal (sometimes a lot!) of money. An interesting fact reported by Davide: the time required between the compromisation of a new host (to drop malicious payload) and the generation of new maldocs pointing to this host is only… 3 hours!
The next presentation was performed by Gal Bitensky ( Minerva):  “Vaccination: An Anti-Honeypot Approach”. Gal (re-)explained what the purpose of a honeypot and how they can be defeated. Then, he presented a nice review of ways used by attackers to detect sandboxes. Basically, when a malware detects something “suspicious” (read: which makes it think that it is running in a sandbox), it will just silently exit. Gal had the idea to create a script which creates plenty of artefacts on a Windows system to defeat malware. His tool has been released
here.
Paul Alderson (FireEye) presented “Injection without needles: A detailed look at the data being injected into our web browsers”. Basically, it was a huge review of 18 months of web-Âinject and other configuration data gathered from several botnets. Nothing really exciting.
The next talk was more interesting… Back to the roots:
SWITCH presented their DNS Firewall solution. This is a service they provide not to their members. It is based on DNS
RPZ. The idea was to provide the following features:
- Prevention
- Detection
- Awareness
Indeed, when a DNS request is blocked, the user is redirected to a landing page which gives more details about the problem. Note that this can have a collateral issue like blocking a complete domain (and not only specific URLs). This is a great security control to deploy. Note that RPZ support is implemented in many solutions, especially Bind 9.
Finally, the first day ended with a presentation by Tatsuya Ihica from Recruit CSIRT: “Let your CSIRT do malware analysis”. It was a complete review of the platform that they deployed to perform more efficient automatic malware analysis. The project is based on Cuckoo that was heavily modified to match their new requirements.
The second day started with an introduction to the FIRST organization made by Aaron Kaplan, one of the board members. I liked the quote given by Aaron:
If country A does not talk to country B because of ‘cyber’, then a criminal can hide in two countries
Then, the first talk was really interesting: Chris Hall presented “Intelligence Collection Techniques“. After explaining the different sources where intelligence can be collected (open sources, sinkholes, …), he reviewed a serie of tools that he developed to help in the automation of these tasks. His tools addresses:
- Using the Google API, VT API
- Paste websites (like pastebin.com)
- YARA rules
- DNS typosquatting
- Whois queries
All the tools are available here. A very nice talk with tips & tricks that you can use immediately in your organization.
The next talk was presented by a Cisco guy, Sunil Amin: “Security Analytics with Network Flowsâ€. Netflow isn’t a new technology. Initially developed by Cisco, they are today a lot of version and forks. Based on the definition of a “flowâ€: “A layer 3 IP communication between two endpoints during some time periodâ€, we got a review the Netflow. Netflow is valuable to increase the visibility of what’s happening on your networks but it has also some specific points that must be addressed before performing analysis. ex: de-duplication flows. They are many use cases where net flows are useful:
- Discover RFC1918 address space
- Discover internal services
- Look for blacklisted services
- Reveal reconnaissance
- Bad behaviours
- Compromised hosts, pivot
- HTTP connection to external host
- SSH reverse shell
- Port scanning port 445 / 139
I would expect a real case where net flow was used to discover something juicy. The talk ended with a review of tools available to process net flow data: SiLK, nfdump, ntop but log management can also be used like the ELK stack or Apache Spot. Nothing really new but a good reminder.
Then, Joel Snape from BT presented “Discovering (and fixing!) vulnerable systems at scale“. BT, as a major player on the Internet, is facing many issues with compromized hosts (from customers to its own resources). Joel explained the workflow and tools they deployed to help in this huge task. It is based on the following circle: Introduction, Â data collection, exploration and remediation (the hardest part!).
I like the description of their “CERT dropbox” which can be deployed at any place on the network to perform the following tasks:
- Telemetry collection
- Data exfiltration
- Network exploration
- Vulnerability/discovery scanning
An interesting remark from the audience: ISP don’t have only to protect their customers from the wild Internet but also the Internet from their (bad) customers!
Feike Hacqueboard, from TrendMicro, explained: Â “How political motivated threat actors attack“. He reviewed some famous stories of compromised organizations (like the French channel TV5) then reviewed the activity of some interesting groups like C-Major or Pawn Storm. A nice review of the Yahoo! OAuth abuse was performed as well as the tab-nabbing attack against OWA services.
Jose Enrique Hernandez (Zenedge) presented “Lessons learned in fighting Targeted Bot Attacks“. After a quick review of what bots are (they are not always malicious – think about the Google crawler bot), he reviewed different techniques to protect web resources from bots and why they often fail, like the JavaScript challenge or the Cloudflare bypass. These are “silent challenges”. Loud challenges are, by examples, CAPTCHA’s. Then Jose explained how to build a good solution to protect your resources:
- You need a reverse proxy (to be able to change quests on the fly)
- LUA hooks
- State db for concurrency
- Load balancer for scalability
- fingerprintjs2 / JS Challenge
Finally, two other Cisco guys, Steve McKinney & Eddie Allan presented “Leveraging Event Streaming and Large Scale Analysis to Protect Cisco“. CIsco is collecting a huge amount of data on a daily basis (they speak in Terabytes!). As a Splunk user, they are facing an issue with the indexing licence. To index all these data, they should have extra licenses (and pay a lot of money). They explained how to “pre-process” the data before sending them to Splunk to reduce the noise and the amount of data to index.
The idea is to pub a “black box” between the collectors and Splunk. They explained what’s in this black box with some use cases:
- WSA logs (350M+ events / day)
- Passive DNS (7.5TB / day)
- Users identification
- osquery data
Some useful tips that gave and that are valid for any log management platform:
- Don’t assume your data is well-formed and complete
- Don’t assume your data is always flowing
- Don’t collect all the things at once
- Share!
Two intense days full of useful information and tips to better defend your networks and/or collect intelligence. The slides should be published soon.
3 comments