Threat Catalogue

Attacker directly modifies cluster state in etcd, bypassing Kubernetes API server admission controllers and RBAC to create privileged workloads, modify RBAC rules, or corrupt cluster configuration

Attacker breaks out of container isolation to access the host system

Attacker accesses etcd snapshots or backups which contain complete cluster state including all secrets, ConfigMaps, service account tokens, and RBAC configurations

Attacker accesses Kubernetes secrets stored in etcd, which may contain credentials, API keys, TLS certificates, and other sensitive data often stored without encryption-at-rest by default

Attacker elevates their access rights to gain higher-level permissions

Adversary encrypts or destroys organisational data and demands payment for decryption. Modern variants combine encryption with double-extortion exfiltration, threatening to publish stolen data even if backups allow recovery. Cloud-targeted ransomware additionally abuses identity to delete snapshots, rotate KMS keys, or wipe object storage.

Attacker injects malicious SQL queries through user input to manipulate database operations

Attacker leverages management tools to execute unauthorized commands on target systems

Attacker gains control of user accounts through credential stuffing, phishing, or session hijacking

Attacker steals authentication tokens to impersonate users or gain unauthorized access

Database backups stored insecurely expose sensitive data

Attacker modifies batch job definitions, scripts, or execution environment to execute malicious code or access unauthorized resources

Weak authentication mechanisms allow attackers to compromise user accounts

Attacker injects malicious code that gets executed by the application

Attackers steal credentials to gain unauthorized access to systems

Unauthorized extraction of sensitive data from the organization

Attacker circumvents data lake access controls, column/row-level security, or data masking policies to access restricted data

Attacker disrupts the Raft consensus protocol by manipulating leader election, injecting rogue members, or causing network partitions that lead to split-brain scenarios or data inconsistency

Adversary takes control of DNS resolution to redirect traffic to attacker-controlled infrastructure for credential capture, traffic interception, or malware delivery. Attack surfaces include compromised registrar accounts, unauthorised changes to authoritative records, resolver cache poisoning, and DNS rebinding attacks that abuse browser same-origin assumptions to reach internal services.

Attacker modifies data transformation logic or job definitions to corrupt data, exfiltrate information, or inject malicious content

Attacker exploits GraphQL-specific vulnerabilities including query depth attacks, introspection abuse, batching attacks, or field-level authorization bypass

Attacker exploits hybrid cloud connections to pivot between on-premises and cloud environments or intercept traffic on dedicated connections

Authorised individuals (employees, contractors, or partners) misuse legitimate access to exfiltrate data, sabotage systems, or commit fraud — either intentionally or through negligence. Distinct from credential theft because the actor already holds valid permissions, which makes detection dependent on behavioural rather than authentication signals.

Attacker moves between systems within the network after initial compromise

Malicious software installed on systems to perform unauthorized actions

Attacker manipulates NoSQL queries through unsanitized input to access or modify data

Attackers exploit OAuth tokens to access connected applications, exfiltrate data, or perform unauthorized actions across integrated services

Adversary uses deceptive emails, messages, voice calls, or look-alike sites to trick users into revealing credentials, executing malicious payloads, or approving fraudulent transactions. Targeted variants (spearphishing, business email compromise) impersonate specific colleagues or vendors and are a primary initial-access vector for cloud account takeover.

Attacker manipulates AI model inputs to bypass safety controls, extract sensitive information, or cause unintended behavior

Attacker tricks server into making requests to internal resources or external systems

Cloud storage buckets configured with overly permissive access policies

Malicious code introduced through compromised dependencies or build processes

Attacker gains access to resources without proper authentication or authorization

Attacker injects malicious content into cache to serve compromised data to users

Attacker structures, encodes, or obfuscates sensitive data to evade automated detection and classification systems

Sensitive data transmitted between services may be exposed if not properly encrypted

Attacker intercepts or manipulates real-time data streams to inject false data or steal information

Adversary forges email sender identity to impersonate trusted parties for phishing, fraud, or business email compromise. Enabled by missing or permissive SPF/DKIM/DMARC records, open mail relays, or look-alike domains; closely paired with phishing as the delivery technique but distinct in that the failure is in mail-authentication policy rather than user judgement.

Attacker injects malicious events into event-driven systems to trigger unauthorized actions

Services or users granted more permissions than necessary for their function

Attacker intercepts and potentially modifies communication between two connected services

Attacker modifies messages in transit between services

Attacker queries AI model systematically to reconstruct or steal proprietary model weights and architecture

Improper network security group or firewall rules expose resources

Attacker injects malicious data into the communication channel between services

Attacker captures and retransmits valid data transmissions to perform unauthorized actions

Sensitive data unintentionally shared externally through SaaS collaboration features such as file sharing, public links, or guest access

Attacker manipulates search indexes to inject malicious content, alter search rankings, or expose sensitive data through search results

Insecure default configurations or incomplete setup creates security vulnerabilities

Sensitive data exposed through improper storage, transmission, or access controls

Unauthorized SaaS applications used by employees bypass security controls, create visibility gaps, and may expose corporate data

Insecure integrations with third-party applications create attack vectors or expose sensitive data through connected services

Attacker manipulates training data or fine-tuning datasets to introduce backdoors or biases into AI models

Known security vulnerabilities remain unpatched, enabling exploitation

Attacker exploits workflow orchestration to bypass security controls or execute unauthorized steps

Attacker disables or evades security logging to hide malicious activity

Attacker overwhelms the connection between services to disrupt communication

Attack aimed at making services unavailable to legitimate users

Attacker modifies or deletes logs to hide malicious activity