Threat Hunting in Practice
By Robert Salier, Product Manager, Endace
Hunting for security threats involves looking for traces of attackers in an organization’s IT environment, both past and present. It involves creativity combined with (relatively loose) methodologies and frameworks, focused on outsmarting an attacker.
Threat Hunting relies on a deep knowledge of the Tactics, Techniques and Procedures (TTP’s) that adversaries use, and a thorough knowledge of the organization’s IT environment. Well executed threat hunts provide organizations with deeper insight into their IT environment and into where attackers might hide.
This, the second article in our series of blog posts on threat hunting (read Part 1 here), looks at how leading organizations approach threat hunting, and the various data, resources, systems, and processes required to threat hunt effectively and efficiently.
Larger organizations tend to have higher public profiles, more valuable information assets, and complex and distributed environments that present a greater number of opportunities for criminals to infiltrate, hide, and perform reconnaissance without detection. When it comes to seeking out best practice, it’s not surprising that large organizations are the place to look.
Large organizations recognize that criminals are constantly looking for ways to break in undetected and that it is only a matter of time before they succeed if they haven’t already. While organizations of all sizes are being attacked, larger organizations are the leaders in this proactive approach to hunting down intruders, i.e. “threat hunting”. They have recognized that active threat hunting increases detection rates over-relying on incident detection alone – i.e. waiting for alerts from automated intrusion detection systems that may never come.
Best practice involves formulating a hypothesis about what may be occurring, then seeking to confirm it. There are three general categories of hypothesis:
- Driven by threat intelligence from industry news, reports, and feeds.
e.g. newsfeeds report a dramatic increase in occurrences of a specific ransomware variant targeting your industry. So a threat hunt is initiated with the hypothesis that your organization is being targeted with this ransomware
- Driven by analytics, i.e. starting with possible Indicators of Attack (IoA’s) and Indicators of Compromise (IoC’s) flagged by internal systems such as the IDS.
e.g. a bank notices a gradual increase in fraudulent credit card transactions, so the hypothesis is that credit cards are possibly being skimmed from POS terminals
- Driven by situational awareness, i.e. focus on infrastructure, assets and data most important to the organization.
e.g. a hypothesis that your customers’ records are the “crown jewels”, so hackers will be trying to gain access to exfiltrate this data
Having developed a hypothesis as a starting point, leading organizations rely on a range of tools and resources to threat hunt efficiently and effectively:
Historic Data from Hardware, Software and the Network
- Infrastructure Logs from the individual components of hardware and software that form your IT environment, e.g. firewalls, IDS, switches, routers, databases, and endpoints. These logs capture notable events, alarms and other useful information, which when pieced together can provide valuable insight into historic activity in your environment. They’re like study notes that you take from a text book, i.e. highly useful, but not a full record, just a summary of what is considered notable. Also, be wary that hackers often delete or modify logs to remove evidence of their malicious activity.
- Summarized network data (a.k.a. “packet metadata”, “network telemetry”). Traffic on network links can be captured and analyzed in real time to generate a feed of summary information characterizing the network activity. The information that can be obtained goes well beyond the flow summaries that Netflow provides, e.g. by identifying and summarizing activity and anomalies up to and including layer 7 such as email header information and expired certificates. This metadata can be very useful in hunts and investigations, particularly to correlate network traffic with events and activity from infrastructure logs, and users. Also, unlike logs, packet metadata cannot be easily deleted or modified.
- Packet level network history. By capturing and storing packets from a network link, you have a verbatim copy of the communication over that link, allowing you to see precisely what was sent and received, with zero loss of fidelity. Some equipment such as firewalls and IDS’s capture small samples of packets, but these capture just a fraction of a second of communications, and therefore must be automatically triggered by a specific alarm or event. Capturing and storing all packets (“full packet capture”, “100% packet capture”) is the only way to obtain a complete history of all communications. Historically, the barriers to full packet capture have been the cost of the required storage and the challenge of locating the packets of interest, given the sheer volume of data. However, recent advances in technology are now breaking down those barriers.
Baselines
Baselines are an understanding of what is normal and what is anomalous.
Threat hunting involves examining user, endpoint, and network activity, searching for IoA’s and IoC’s – i.e. “clues” pointing to possible intrusions and malicious activity. The challenge is knowing which activity is normal, and which is anomalous. Without knowing that, in many cases, you will not know whether certain activity is to be expected in your environment, or whether it should be investigated.
A Centralized Location for Logs and Metadata
Because there are so many disparate sources of logs, centralized collection and storage is a practical necessity for organizations with substantial IT infrastructure. Most organizations use a SIEM (Security Information and Event Manager), which may have a dedicated database for storage of logs and metadata, or may use an enterprise data lake. SIEMs can correlate data from multiple sources, support rule-based triggers, and can feature Machine Learning algorithms able to learn what activity is normal (i.e. “baselining”). Having learned what is normal, they can then identify and flag anomalous activity.
Threat Intelligence
Threat intelligence is knowledge that helps organizations protect themselves against cyber attacks. It encompasses both business level and technical level detail. At a business level this includes general trends in malicious activity, individual breaches that have occurred, and how organizations are succeeding and failing to protect themselves. At a technical level, threat intelligence provides very detailed information on how individual threats work, informing organizations how to detect, block, and remove these threats. Generally this comes in the form of articles intended for consumption by humans, but also encompasses machine-readable intelligence that can be directly ingested by automated systems, e.g. updates to threat detection rules.
Frameworks and Regulations
The regulatory environment influences threat hunting, and cyber defense in general. In many countries, regulations impose obligations on disclosure of breaches, including what information must be provided, when, and to which stakeholders. There are a also a number of frameworks addressing cyber security at the governance level, which in some cases overlap with regulations, dealing with many of the same issues and considerations. Collectively, these frameworks and regulations help to ensure organizations implement good strategies, policies, processes and tools.
In the next article in this series, we explore the frameworks and regulations that apply to threat hunting, and which ensure organizations implement appropriate strategies, policies, processes and tools.