All users who handle electronic devices are given an identity. If the name of a human is considered their ‘Identifier’, then the username and password, biometric login, physical authentication key, etc. which they use to authorize and authenticate into systems, applications and other devices are all examples of their ‘identity’. This human identity helps certify anyone who accesses or connects to a machine.
The growth of cloud services, automation, and reliance on digital workflows have introduced the creation of various machine accounts or ‘cyber identities’. They are used to simplify complex and repetitive workflows that would otherwise require continuous human interaction/intervention. They can also streamline certain tasks to increase operational efficiency.
Industrial OT leverages service accounts for their controllers and completely automate data collection, machine operations and robotic processes (RPA) with minimal to no human intervention.
IoT aims to create a world where physical utilities seamlessly integrate into the IT environment to participate in the business. This includes both consumer electronics like smart devices, security cameras etc. and industrial devices like embedded sensors and actuators that use NHIs to communicate with one another.
We have moved far ahead from simple AI agents such as chatbots and co-pilots, these early monolithic systems had low autonomy and heavily relied on human input. ‘Digital AI workers’ or Agents now construct their own instructions, augmenting human input and becoming autonomous to a stage where they can create more non-human identities themselves to creatively carry out their tasks.
Non-human identities can be defined as the unique cyber identities used by applications, servers, and other devices to certify themselves when connecting to other systems and services. Non-human identities create connections to bridge the gap between business that have multiple disparate environments in the hybrid-cloud era.
Their primary function is to enable machine-to-machine access within a software ecosystem. Non-human identities are also known as ‘non-carbon’ users or ‘machine identities’ to represent their digital presence.
NHIs are vital to mimic actions that a human would have to perform multiple times or actions that are highly complex and cannot afford errors. A few examples of this are ‘consent fatigue’: the need to constantly approve permissions for applications and processes, and ‘data manipulation’: operations like filtering, sorting, merging, aggregating, and transforming data. In both these cases, NHIs are used - Oauth Service for consent automation Accounts and RPAs for data operations.
In most cases, research articles and surveys portray machine identities and non-human identities to be one and the same. However, machine identities generally refer to the wide range of identities associated with systems, and software. And non-human identities encompass the overall governance and visibility of these machine identities, CIEM tools, SaaS applications and service accounts and so on.
Non-human identities are spread across the enterprise IT environment to support and enable various operations. In a typical digital ecosystem, they are prevalent across cloud services (Google, Azure, and AWS), Code (CI/CD pipelines), SaaS applications, and a few hosted resources. A representative diagram of the distribution is shown below.
In cloud environments, NHIs are crucial to achieve a solid cybersecurity strategy, helping decrease business risk and security breaches significantly.
In CI/CD environments, NHIs are generally used to automate data processing and manage critical services. In Kubernetes for example, these NHIs rely on IAM roles, assigned through machine-specific tags and external variables, to retrieve credentials from secret stores. An example of a typical distribution of NHIs is as follows:
Software applications and APIs are often assigned their own identities to interact with other systems securely. These identities typically use API keys or OAuth tokens for authentication.
1. Non-human Identity Proliferation
New identities are constantly being created for automation and there is minimal to no governance over them. Engineers and devs also consistently add and create secrets that connect services within the internal infrastructure to enable automated processes. This results in a spill of non-human identities, making it difficult to gain sight of each of them.
The ratio of human identities to machine identities averages from 1:50 to a 1:100 ratio and is only expected to increase with technological advancements in cloud (SaaS) services.
2. Lack of Traceability
With the increasing number of NHIs, companies often lose track of the number of non-human identities being created. This makes it difficult to govern them and creates a new atttack surface – creating gaps for malicious actors to misuse.
If, and rather - when a mishap occurs, it becomes impossible to point out which NHI was responsible for the incident. This lack of transparency and traceability is a huge problem for incident response and forensics operations as they cannot hold anyone accountable or retrace steps to minimize the impact of an attack. They are forced to shut down systems and servers entirely to stop the propagation of damage.
3. Excessive Access Permissions
NHIs, overall, may also have more access permissions than human users. Let's take for example, a service account associated with an application. This service account (an NHI) uses an API key to authenticate into a cloud service like AWS to carry out tasks such as fetching AWS users with the IAM permissions it holds.
The access provisioned to this service account may be more permissive than a human identity as the access is often permanent owing to the continuous and autonomous operating nature of the NHI.
4. Loopholes in AI Threat Detection
Traditional AI anomaly detection tools help stop cyberattacks by analyzing user behavior. Human behavior is predictable to an extent, and a deviation can be derived from simple variables such as a user having an extended access timeframe.
When it comes to non-human identities however, a more advanced approach is required for risk detection. NHI’s are dynamic when compared to their human counterparts, meaning that they might have frequent changes in their access requirements. This creates difficulties in defining the scope of NHI access that AI will leverage to detect anomalies. Additional factors such as - data retrieval outside the set schedule, privileged access beyond NHI responsibilities etc. need to be taken into consideration.
Ultimately, the scale and complexity of these new entities results in an identity explosion and non-human identities may replace humans to be the weakest link in the security chain if they are not handled properly.
With the expansion of non-human identities, a strong requirement arises to manage them. With NHI’s constantly interacting with an enterprise’s most critical data and systems, they also need to be properly tracked and secured. Without proper tracking and management, hackers may exploit the large and ungoverned attack surface and utilize permissions that NHIs carry to carry out cyberattacks.
Non-human identity management (NHIM) involves identifying the number of NHIs within the software ecosystem, the various types of NHIs involved, permissions they carry and then holistically managing them.
The management lifecycle of non-human identities typically involves:
The classification of these identities as simply ‘machine’ identities make the approach to handling them very complex. A more intricate, well-defined split up will help with classification to understand security needs and curated management of these identities. This is especially beneficial as each NHI may have a different lifecycle, permissions, and associated attributes.
Non-human identity data models often extend beyond traditional Identity and Access Management (IAM) schemas. They often have attributes such as:
Breaking down the major differences between ‘carbon’ and ‘non-carbon’ identities helps implement a solid identity management strategy that addresses the complexities in handling them.
A) How they are created: Human identities are generally manually created by a human user – the IT or security team and do not drastically grow. Meanwhile, NHIs are created in multiple ways, across multiple platforms by various users and applications – proliferating quickly with code and cloud adoption.
B) How they are managed: Human users are manually provisioned, and HR systems can be used to manage them. NHIs are sporadically and continuously created by developers and systems to enable automation and are almost never managed properly.
C) How they are tracked: Human users often have accounts tied to them, generally associated with their ‘username’. This makes it easy to track and trace the activities they carry out. Therefore, each human user is held accountable for their actions. Unlike human identities, NHIs are not associated with specific users, thus getting away from scrutiny by regulations and often used by multiple admins or applications.
D) How they are authenticated: Authentication of human identities relies on these three factors to be secure
NHIs support multiple authentication methods, reflecting technological evolution. Various systems may employ different authentication methods, leading to a wide range of approaches in use.
With NHIs, the only protection is the secret that the user (in most cases a developer) gave to the machine - there is no SSO or MFA in the middle. This means that if attackers get hold of a service account and the secret there isn’t anything else that can stop them.
The compromise of a single NHI can precipitate widespread disruptions, which leads to outages for all applications that rely on them. Unlike human accounts, which typically affect access of an individual user, NHIs have inordinate access rights that can incapacitate multiple devices all at once. This amplifies the potential damage and convolutes recovery efforts following a security incident, thereby necessitating heightened security protocols and protective measures for NHIs.
While organizations partially understand the risk of NHIs, they have not fully committed to securing them. Protecting these identities begins with staying in line with security best practices. At a high level, NHIs can be secured by following these best practices.
1) Identifying and gaining visibility over non-human identities
To have an overview of all NHI’s, the primary approach would be to have a secure inventory of all digital identities within the IT infrastructure. This serves as the base layer for understanding the scope of non-human identities and managing them. Regularly scanning for secrets helps with having a bird's eye view over them and detect anomalies.
2) Managing sensitive non-human secrets
To holistically manage NHIs, sensitive secrets need to be handled from the secure inventory. They need to be monitored from creation to expiry and assigned an owner for better accountability and tracking.
3) Applying security and credential best practices
Keeping the security policies consistent across all identities is crucial for enterprise security. Policies must be enforced to frequently rotate credentials, regularly monitor them and de-provision them when no longer in use
While organizations have fortified human user access with robust security measures, the management and security of non-human access like RPA bots, service accounts, API keys and tokens have not received the same level of scrutiny. NHIM brings a paradigm shift in privileged access management (PAM). While identity has become the new security perimeter, maintaining focus on human identities is no longer enough. Organizations need specialized solutions designed specifically for the unique requirements of non-human entities to address fundamental requirements. Unified PAM helps secure NHIs by:
1) Providing holistic visibility: PAM helps provide visibility into NHIs and service accounts spread across hybrid cloud environments, including AWS, Azure, and Google Cloud.
2) Vaulting and secure NHIs: PAM provides a secure way to store DevOps secrets, keys and certs in an encrypted vault. It also allows developers to leverage APIs to support automation in developer environments.
3) Assessing security: Unified PAM can actively assess the security posture of non-human identities, check for compliance and provide reports.
4) Managing the NHI lifecycle: PAM can track the creation and expiration dates of SSH keys, and certificates, enhancing operational efficiency and security.