“We log everything” is a sentence I would treat as a warning. Usually it means one of three things: the logs exist but nobody can search them, they exist but do not contain the field you need, or they are about to be overwritten. Let us take the common beliefs one at a time.

Belief: more logs mean more security

Volume is not value. Storage and SIEM licences are often priced by data ingested, so recording every debug message from every application is expensive and buries the useful lines. Logging should be driven by questions you will need to answer after an incident: who logged in, from where, what did they access, what did they change, what ran on the machine, what left the network?

A reasonable priority list of sources:

  • Authentication and identity events, including successes, failures, MFA outcomes, password resets and changes to privileged group membership.
  • Endpoint and server events, notably process creation and command-line details, which are what reveal attacker tooling.
  • Network edge: firewall, proxy, VPN and DNS logs, which show where internal machines talk.
  • Cloud control-plane activity, meaning the record of every administrative API call.
  • Email gateway and mailbox audit logs.
  • Application audit trails for sensitive business actions, such as exporting customer data or approving a payment.

Belief: default settings will do

Many systems do not record the useful details unless configured to. Windows, for example, does not log command lines of new processes by default; you must enable the relevant auditing policy. Cloud accounts sometimes keep only a short window of activity unless you create a trail to longer-term storage. Web servers may log requests without recording the real client address when a load balancer sits in front. Check your key sources by running a harmless test action and seeing whether you can find it afterwards. This single exercise exposes more gaps than any policy document.

Belief: the timestamps will line up

They will not, unless you make them. When investigating an intrusion you stitch together events from a firewall, a domain controller, a laptop and a cloud service. If one device’s clock is eight minutes off, the sequence becomes nonsense, and you may blame the wrong machine. Synchronise everything to a trusted time source using NTP, and prefer recording timestamps with a time zone, ideally UTC internally. Indian organisations also note that the Indian government’s CERT-In directions of 2022 require service providers, intermediaries, data centres, body corporates and government organisations to sync system clocks to the NIC or NPL NTP servers or sources traceable to them. Verify the current text of the directions for your sector.

Belief: keep logs as long as the disk allows

Retention should be a decision. Attackers often sit inside a network for weeks before being found, so thirty days of logs may start after the break-in. Many organisations keep recent logs “hot” and searchable for around 90 days and older ones in cheaper archive storage for a year or more. The same CERT-In directions mention maintaining logs of ICT systems for a rolling 180 days within Indian jurisdiction, though again, confirm the applicable rule for your organisation. Other drivers include contracts, sector regulators and standards like PCI DSS, which has its own log retention expectations. Longer retention also means more personal data held, so under the DPDP Act, keep only what you can justify.

Belief: logs are trustworthy by nature

An intruder with administrative rights will try to clear or alter local logs. Forward logs to a separate collector as events happen, restrict who can delete them, and consider write-once storage for critical records. Also avoid logging secrets: passwords, tokens, full card numbers and Aadhaar numbers have no place in a log file, and leaving them there turns your log store into a prize.

A final test

Pick a plausible incident: a user’s account was used from abroad last Tuesday at 3 a.m. Could your logs tell you which account, which session, what was opened, whether data was downloaded, and whether the same attacker touched other systems? Wherever the answer is “no”, you have a logging backlog, and that list is more useful than any inventory of what you currently collect.

Leave a Reply

Trending