Access control is the set of rules that decides which station, user, or device may use a shared resource, and the choice between DAC, MAC, RBAC, and ABAC is driven by user count, audit depth, and latency budget, not by marketing.
For industrial buyers the practical spectrum runs from a 2-station TDMA-style fixed-slot gate (collision-free, predictable, but ~30-50% slot waste under bursty load [S1]) to a full Role-Based Access Control (RBAC) scheme with policy, role, and permission layers. The Chinese journal paper An Enhanced RBAC Scheme in Classified Information System (Vol. 13, Issue 11) explicitly compares DAC's per-object granularity with RBAC's easier management, then proposes a hybrid that retains DAC detail while inheriting RBAC administration [S2]. That hybrid is the conceptual ancestor of most modern attribute-aware door controllers.
TDMA-style fixed-slot access: where it still wins
TDMA divides a shared channel into round-robin time slots; each station transmits only in its own slot, so the design is collision-free by construction and needs minimal access-control logic [S1]. On the downside, when a station has nothing to send, its slot is wasted, all stations must hold tight clock sync, and a low-priority node still waits its full slot, so fixed-slot schemes are a poor fit for bursty or real-time traffic [S1]. For a 2-4 node industrial sensor bus with predictable poll rates, the simplicity of fixed-slot access still beats CSMA's collision-retry overhead.
RBAC vs DAC vs MAC vs ABAC: a 4-criteria selection table
Specifying an access control system means choosing between four canonical models, each with a different answer to the same four questions. The table below lines them up against the criteria that drive industrial purchase orders: where the decision lives, how roles scale, typical user counts, and audit trail depth. The Chinese classified-system paper concludes that pure DAC offers the finest per-object control but explodes administratively past a few hundred subjects, while pure RBAC collapses granularity for the sake of simple role assignment, and the optimal answer is a layered model [S2].
Comparison of the four dominant access-control models on decision criteria that matter to a spec engineer:
- DAC (Discretionary): owner decides who gets in; high per-object granularity, low admin scalability; fits <200 users; weak centralized audit, since policy lives with each data owner [S2].
- MAC (Mandatory): classification labels (e.g. Confidential / Secret) drive decisions, users cannot override; high assurance, but rigid; fits defence, government, classified networks as analysed in [S2].
- RBAC (Role-Based): permissions attach to roles, users attach to roles; scales to 10,000+ users with role hierarchies and separation-of-duty constraints; this is the model the [S2] paper extends with classified-system attributes.
- ABAC (Attribute-Based): policy engine evaluates user, resource, and environment attributes together; most flexible, most expensive in policy-engine compute, fits multi-tenant cloud and OT/IT converged plants.
A useful rule of thumb from the [S2] analysis: pick RBAC when the workforce is large and role churn is low, pick ABAC when tenant, shift, and clearance attributes must all be evaluated per request, and pick MAC only when regulatory classification (e.g. national-security grade) is the binding constraint.
Physical and electronic access control: card, biometric, and two-hand
On the factory floor, access control often means a door controller, not a software role engine. Card-only readers (Wiegand 26/34-bit, 13.56 MHz MIFARE DESFire EV2, or 125 kHz EM4100) cost the least per door but expose the system to lost-card and tailgating risk. Adding biometric (fingerprint template-on-card, or 1:N face match at 0.5-1 s per attempt) closes that gap, at the cost of a 2-4x higher per-door bill of materials and stricter GDPR/CCPA template storage rules. [S1]
For high-hazard machinery, two-hand control is the relevant standard: EN 574 Type III-C requires two actuators operated within 500 ms of each other, both held during the hazard, and a release-and-repress to restart, which prevents one-handed defeat. Pairing a two-hand station with an access control badge ensures only authorized, trained operators can energize the press, while the fire alarm control panel drops the mag-lock on general evacuation.
Integration with industrial subsystems: ASRS, valves, and cable choices
Inside an ASRS system, access control governs which operator may issue a put-away or retrieval command, and the control cable carrying those commands is usually shielded twisted pair rated for continuous flex (e.g. LiYCY 4x0.34 mm² or Profinet Type C, drag-chain rated to 5-10 million cycles). A control valve on a hazardous line typically interlocks with both the access-control badge and a position-proof switch, so the valve cannot be stroked unless a permitted role is logged at the HMI within the last 30 s; this is the same "subject + action + freshness" pattern that the [S2] paper's enhanced RBAC scheme enforces in software.
Failure modes and engineering limits you must spec for
Three failure modes recur across all four access-control models and are worth pinning in the spec. First, role explosion: an RBAC deployment that grows past 200-400 distinct roles without a role-mining pass becomes harder to administer than the DAC mess it replaced, a finding consistent with the [S2] paper's motivation for hybrid models. Second, clock and sync dependence: TDMA-style fixed-slot schemes need synchronization typically held within ±1-5% of slot time, otherwise slot overlap causes the very collisions the design was meant to eliminate [S1]. Third, audit-tell-tale gap: many physical access-control installs log badge reads but not tailgate events, so a forced-door event passes unrecorded; biometric or stereo-vision turnstiles close that gap at higher per-door cost.
Procurement signals worth tracking
Two signals are worth watching through 2026. (1) Hybrid RBAC/ABAC deployments, the model the [S2] paper proposed for classified systems, are appearing in OT/IT converged plants where role-only RBAC cannot express shift + clearance + location in one decision. (2) For greenfield sites, vendors are increasingly pairing card readers with native OSDP (Open Supervised Device Protocol) v2.2 wiring, which encrypts the reader-to-controller link and removes the Wiegand plaintext exposure that most legacy access-control panels still carry. For deeper TCO data on the camera side of physical access, see this industrial surveillance camera TCO breakdown and for adjacent vision-system sourcing, the machine vision system supplier tier map lays out how spec filters cluster across 2026 vendors.