Toll Cameras Are Now Police Cameras: NSW's Privacy Shift and What Drivers Can Do

The Threshold Just Lowered

If you drive through Sydney on any given day, you pass under a toll camera roughly once every few kilometres. Until now, those images were collected by private toll road operators like Transurban, and when police requested them, the images were redacted — faces blurred, passengers obscured, identifying details stripped.

That’s changing.

The NSW Minns government has introduced legislation that will give NSW Police access to unredacted toll camera images when investigating serious indictable offences, missing persons cases, or coronial matters. Phase one targets the Sydney Harbour Bridge and Harbour Tunnel — the busiest corridors in the state — with a broader rollout to follow.

Current System	Proposed System
Images redacted (faces, details obscured)	Full, unredacted images
Limited investigative utility	Identification of driver, passengers, behaviour
Collected by private operators	Shared directly with law enforcement
Used for toll enforcement only	Used for serious crime, missing persons, coronial

The Police Minister, Yasmin Catley, described it as giving officers “another important investigative tool” — part of a plan to “give our police the powers and capability they need to keep communities safe.”

On its face, that sounds reasonable. Who would argue against solving murders or finding missing people?

But the real story isn’t about the edge cases where this helps catch a killer. It’s about the surveillance infrastructure this normalises — and what happens to your privacy once the precedent is set.

What Toll Cameras Actually See

Toll cameras aren’t simple ANPR (Automated Number Plate Recognition) units. Modern toll gantries carry multi-lens camera arrays that capture:

• The vehicle’s front and rear number plates (day and night, all weather)
• The driver’s face through the windscreen
• Passengers in the front and rear seats
• The vehicle’s make, model, colour, and modifications
• The date, time, lane, and direction of travel
• Speed at the moment of capture

These aren’t blurry CCTV grabs. Toll cameras are positioned close to traffic, at fixed angles, with high-resolution sensors tuned to read plates through glare, rain, and darkness. The image quality is deliberately high because the system needs to reliably bill the correct vehicle owner.

Here’s what this means: every trip through a toll point generates a forensic-quality photograph of you, your car, your passengers, and your location at a precise moment in time.

Under the current (redacted) system, police could access metadata — “plate XYZ-123 crossed the bridge at 14:32 on Tuesday” — but not the image itself. Under the proposed system, they get the picture. And with it, the ability to:

• Confirm who was driving (not just who owns the car)
• Identify passengers and their relationships
• Establish behavioural patterns (phone in hand? eating? looking down?)
• Cross-reference faces against police databases (wanted persons, persons of interest)
• Build timelines of multiple vehicles across multiple toll points

The Expectation of Privacy on a Public Road

Australian law has historically taken a dim view of privacy on public roads. The High Court’s reasoning in ABC v Lenah Game Meats (2001) and subsequent ANPR-related cases has generally held that there’s no reasonable expectation of privacy in something observable from a public space.

But there’s a gap between “someone could see you” and “the government maintains a searchable, permanent record of every vehicle’s movements with unblurred faces.”

The UK offers a cautionary tale. The UK’s ANPR network, operated by the National ANPR Service, processes over 50 million number plate reads per day. Data is retained for up to two years. The system has been criticised by the UK’s Biometrics Commissioner for function creep — originally sold as a tool for tackling vehicle crime, it’s now routinely used for:

• Traffic offence enforcement (speeding, bus lane violations)
• Insurance validation (cross-referencing plates against insurance databases)
• Intelligence gathering (building movement profiles of non-suspects)
• Child support enforcement (locating parents who owe payments)
• Local council enforcement (tracking vehicles that enter clean-air zones)

The Australian Privacy Foundation has long warned that ANPR systems are “honeypots” that attract scope creep, writing: “The database can be used to draw inferences and generate suspicions… the database is a ‘honeypot’ that attracts attention from many organisations for many purposes, resulting in ‘scope creep’.”

The NSW government says this power will be limited to “serious indictable offences” — a category that includes offences punishable by 5+ years imprisonment. But legislative safeguards have a history of being quietly broadened. Once the infrastructure is in place and the redaction process removed, the incremental cost of expanding access is near zero.

What Drivers Can Do to Protect Themselves

If you want to limit the surveillance surface of your vehicle, you have options — ranging from the simple to the technically sophisticated.

1. Reduce Your Toll Road Exposure

The most effective measure is also the simplest: use alternative routes. If your commute doesn’t require a toll road, don’t use one. Every toll gantry you pass adds an entry to a permanent, image-verified database of your movements.

• Map apps (Google Maps, Apple Maps) let you set “avoid tolls” as a default preference
• For occasional toll road use, your exposure is minimal; for daily commuters, the data density is high enough to reconstruct your entire weekly schedule

That said, in Sydney, avoiding tolls often means significantly longer travel times. This isn’t a practical solution for everyone.

2. Block the Camera’s View of the Cabin

Toll cameras photograph through the windscreen at an angle. You can reduce what they capture by:

• Rear sunshades or curtains — not just on the back window, but on rear side windows where passengers sit. Retractable shades give you the option to lower them when not under cameras.
• Dashcam windscreen mounting — a dashcam mounted high on the windscreen can partially obscure the driver’s face from the camera’s angle without blocking your own view.
• C-pillar positioning — adjusting your seat so the vehicle’s A-pillar or rearview mirror mount falls between you and the roadside camera can break the clear line of sight.

These are not silver bullets. A windscreen is a large target, and multiple cameras on the same gantry capture different angles. But they increase the uncertainty and degrade the quality of identification.

3. License Plate Manipulation (Proceed with Caution)

I need to be clear here: deliberately obscuring your number plate on public roads is illegal in all Australian states and territories. The relevant laws carry fines and, in some cases, demerit points.

That said, the adversarial perturbation community has developed techniques that work within the legal grey area:

Technique	Legal Risk	Effectiveness
IR-blocking plate covers — tinted covers that appear clear to the human eye but block near-infrared illumination used by many ANPR cameras	Likely illegal (NSW Road Rules 2008, Reg 210) in practice	Moderate — defeats IR-only cameras, but most toll cameras use visible-light backup
Adversarial plate surrounds — printed patterns around the plate that exploit OCR model vulnerabilities (e.g., adding lines that confuse character segmentation)	Grey area — not yet tested in Australian courts	Low to moderate — works on older ANPR systems; modern multi-frame systems are more robust
UV-reflective treatments — coatings that cause plate characters to wash out under high-intensity flash	Likely illegal (obscuring plate characters)	Low — toll cameras use controlled exposures that compensate for reflectivity
Standard legal plates in good condition — ironically, faded or dirty plates that are technically legal are harder for ANPR to read than brand-new reflective plates	Fully legal	Low — systems with built-in image enhancement still read them

The adversarial plate approach is worth understanding even if you don’t use it, because it illustrates the core insight of adversarial perturbation: machine learning models see the world differently than humans do. A pattern that looks like a decorative border to you can be a catastrophic distraction to a CNN-based OCR pipeline.

The adversarial perturbation approach illustrated here is conceptually similar to how adversarial patches work against facial recognition: subtle patterns that are imperceptible (or decorative-looking) to humans but systematically degrade the target ML model’s confidence.

4. Trip Fragmentation

One of the most powerful surveillance capabilities is trip reconstruction — linking multiple toll point crossings to build a complete journey. You can disrupt this by:

• Parking and changing drivers at intermediate points (breaks the driver identification chain)
• Exiting and re-entering the toll road network at different points using cash/card payment on non-toll segments
• Using different vehicles for different parts of your regular journey

This mirrors the “session fragmentation” approach from digital tracking: break the continuous data stream into unlinkable segments.

5. Vehicle Choice and Privacy

Not all vehicles are equally trackable. The toll camera database, when cross-referenced with registration data, can identify the registered owner of every vehicle that passes. If you drive a vehicle registered to you personally, every trip is directly linkable to your identity.

Options include:

• Company vehicles registered to a business (adds a layer between the plate data and personal identity)
• Car-share or rental vehicles for specific trips (the association is temporary and shared across many users)
• Registered to a trust or entity rather than an individual (adds administrative friction to any law enforcement inquiry)

This is not about hiding from law enforcement — it’s about controlling the granularity of data available about your personal movements. The question isn’t whether you have something to hide. It’s whether you consent to every trip you take being permanently logged, photographically verified, and available for retrospective analysis.

The Broader Picture

The NSW toll camera change is one piece of a much larger surveillance expansion happening across Australia:

Program	Scope	Status
NSW toll cameras (this bill)	Sydney Harbour Bridge + Tunnel initially, then statewide	Proposed legislation
National ANPR capability	Nationwide plate recognition network	Operational (various jurisdictions)
Victoria’s traffic camera network	Red-light, speed, and point-to-point cameras	Expanding with AI analytics
Queensland’s ANPR expansion	Mobile and fixed cameras	Ongoing rollout
Commonwealth Biometric Matching Capability	Face matching across government databases	Operational since 2022

The toll camera change is significant because it bridges two previously separate systems: administrative toll collection and criminal investigation. Once that bridge exists, the data flows in both directions. Images collected for billing purposes become evidence. Systems designed for traffic management become surveillance tools.

The Australian Privacy Foundation’s position, established since their 2008 ANPR submission, is that ANPR system designs should use a “blacklist in camera” model — where only plates actively being sought are flagged, and all other data is discarded at the point of capture, never stored. This stands in direct opposition to the current trajectory of storing everything and finding uses for it later.

What You Should Actually Do

If you’re an average driver in NSW, here’s the practical checklist:

• [ ] Set your maps app to avoid tolls by default — every avoided gantry is a data point not collected
• [ ] Use retractable rear window shades if you regularly carry passengers
• [ ] Be aware of where toll cameras are on your regular routes — the Harbour Bridge, Harbour Tunnel, M2, M4, M5, Eastern Distributor, Cross City Tunnel, and Lane Cove Tunnel all have them
• [ ] Don’t assume electronic toll tags give you any privacy — the tag ID, timestamp, and location are already collected regardless of the image policy
• [ ] Stay informed about scope creep — the question isn’t whether this power will expand, but when and to what. Subscribe to the Australian Privacy Foundation’s updates

The uncomfortable truth is that once the infrastructure is built and the precedent is set, the privacy battle is largely lost. The most effective countermeasure — not building the system in the first place — was never on the table. What’s left is understanding the system, making informed choices about your exposure, and arguing for legislative safeguards before the next expansion.

“Just because they can see you, doesn’t mean they should be able to find you later.”