Weak-Unforgeable Tags for Secure Supply Chain Management

Abstract

Given the value of imported counterfeit and pirated goods, the need for secure supply chain management is pertinent. Maleki et al. (HOST 2017) propose a new management scheme based on RFID tags (with 2–3K bits NVM) which, if compared to other schemes, is competitive on several performance and security metrics. Its main idea is to have each RFID tag stores its reader events in its own NVM while moving through the supply chain. In order to bind a tag’s identity to each event such that an adversary is not able to impersonate the tag’s identity on another duplicate tag, a function with a weak form of unforgeability is needed. In this paper, we formally define this security property, present three constructions (MULTIPLY-ADD, ADD-XOR, and S-Box-CBC) having this security property, and show how to bound the probability of successful impersonation in concrete parameter settings. Finally, we compare our constructions with the light-weight hash function PHOTON used by Maleki et al. in terms of security and circuit area needed. We conclude that our ADD-XOR and S-Box-CBC constructions have approximately $1/4-1/3$ of PHOTON’s total circuit area (this also includes the control circuitry besides PHOTON) while maintaining an appropriate security level which takes care of economically motivated adversaries.

A NVM-Based RFID Scheme

The NVM-based scheme of Maleki et al. [2] implements the following steps:  

Initialization RFID tag.:

The NVM of the RFID tag is initialized with a sequence of keys $(k^1,k^2,\ldots , k^u)$ and a pointer $p=1$. The back-end server stores the RFID identity ID together with the sequence of keys.

Initialization Reader.:

Each RFID tag reader is initialized with its own key K. The back-end server stores the reader identity together with K.

Reader Event.:

The RFID tag is read out by a reader of a supply chain partner:

1.:

The RFID tage transmits its ID to the reader.

2.:

The reader creates a message m which has the reader identity and time stamp of the event.

3.:

The reader computes $x=MAC_K(m, ID)$. This binds the reader to the event.

4.:

The reader transmits (m, x) to the RFID tag.

5.:

The RFID tag receives (m, x). This triggers the RFID tag to read a next key $k^p$ from its NVM and to increment pointer p by 1. Key $k^p$ is large enough in order to be split up into a first part $k^{p,0}$ and a second part $k^{p,1}$. The tag computes the pair $y^p= (m\oplus k^{p,0}, F_{k^{p,1}}(x))$, where F is a software unclonable function. Since $k^p$ is unique to the RFID tag, the use of function F binds the RFID tag to the event. The key part $k^{p,0}$ serves as a one time pad which prevents traceability.

6.:

The RFID tag stores $y^p$ at the spot where $k^p$ was stored in NVM.

Exit.:

When the tag exits the supply chain, its NVM is read out and communicated to the back-end server. I.e., the internal logic only allows NVM to be read out up to but not including the address pointed at by pointer p. This means that the back-end server receives ID together with $(y^1, \ldots , y^{p-1})$. The ID is used to look up the sequence of keys corresponding to the tag. For each y, this allows the server to first reconstruct the messages m, second to extract the corresponding reader identity with key K from its database, third to compute the mac value $x=MAC_K(m,ID)$, fourth to evaluate F on x with the appropriate key, and finally verify that this is part of y. If all checks pass, then the recorded reader events were not impersonated and they can be verified to correspond to a legitimate path through the supply chain. The server will invalidate the tag for future use in its database.

  A detailed explanation, security analysis, and discussion around how to make the scheme reliable with respect to miss reads and miss writes can be found in [2].

A comparison of state-of-the-art schemes over a range of metrics can be found in [18]. Besides being unique in that, unlike any other scheme, the need for persistent online communication or local databases is avoided, the NVM-based scheme also compares well with most competitive other schemes. The only dimension on which the NVM-based scheme scores negatively is its lack of being able to resist physical attack (where a strong adversary attempts to circumvent the read and write interface in order to clone all the keys stored in NVM). We notice that even though the trace based scheme [8] can withstand physical attacks, the scheme cannot distinguish between a fake and legitimate tag which possibly results in significant financial loss. Current PUF based schemes [19,20,21] are not secure against physical attack because of recent machine learning modeling attacks [22,23,24] – however, as soon as improved PUF designs will resist these modeling attacks, PUF based schemes will resist physical attacks as opposed to the NVM-based scheme. Inherent to current PUF-based schemes, they do need persistent online communication. Also an improved PUF design will likely lead to a higher gate count than the 500–1000 GE for current PUF-based schemes – and this is where the NVM based scheme performs better as well.

As a final note, Sect. 6 discusses several upper bounds on the collision resistance. Obviously, if the resistance is set to $2^{-32}$ or $2^{-64}$, then creating a cloned or fake RFID tag which successfully passes the supply chain becomes very unlikely. In fact too many counterfeit products labelled with fake RFID tags are needed in order to be successful and this makes such an attack economically infeasible. In the introduction we state “An adversary who can circumvent the interface circuitry by means of a physical attack is not considered.” Clearly, the weak link in the NVM-based scheme for high collision resistance will now be its lack of resistance against physical attack.