1 Introduction

Internet of Things (IoT) is an architecture to connect several devices to Internet to manage them or provide different services over them, e.g., to authenticate devices through a cloud server. All computers and also all objects of various elements tend to enter cyberspace and exchange information with each other and this could threaten their security and privacy. Hence, one of the important conditions to the things entrance to the world of the Internet is providing security. Some of the modern cryptographic solutions for providing cyberspace security have been expressed in  [13]. In many IoT architectures, RFID tags are an essential part of them, where they are attached to an object to identify it. To identify an object in a secure way, we need a secure authentication protocol. However, most of those tags are passive and standard authentication protocols, based on asymmetric cryptosystems such as RSA [17] or symmetric cryptosystems such as AES [9], may not be applicable. On the other hand, employing a protocol that does not provide enough security will compromise the user’s privacy. To address this emergence, several protocols have already been proposed in the literature [3, 12, 16]. Among them a type of protocols, that is called ultralightweight protocols, are sound to be more suitable for passive tags. An ultralightweight protocol generally uses a few bitwise operations while computes the messages that are transfered over the protocol. Designing such protocols have a long history on RFID literature, e.g., Gossamer [14], SASI [8] and RAPP [21]) are just examples. Despite these attempts, past studies such as [1, 2, 4, 5, 7, 10, 11, 15] show that it may not be possible to design a secure authentication protocol without employing a secure cryptographic primitive. On the other hand, very recently Tewari and Gupta [20] proposed another ultralightweight authentication protocol to be employed in IoT. The designers have compared the security of their protocol with several other ultralightweight authentication protocols such as Gossamer, SASI and RAPP and claimed that their protocol is secure against desynchronization, secret disclosure and traceability attacks [20, Table 1, Page 15].

Table 1 Notations used in this paper
Full size table

In this paper, we study the security of this protocol and show that, similar to other ultralightweight protocols, this protocol also does not provide desired security against the mentioned attacks. More precisely, we present a very efficient passive attack that retrieves all secret parameters of the tag by only eavesdropping a session of protocol between the target tag and the legitimate reader. The computational complexity of the attack is negligible and can be executed in a fraction of second (we verify our attack by simulation). Our attack ruined any security claim.

1.1 Paper organization

Tewari and Gupta authentication protocol is described in Sect. 2. In Sect. 3, we show how an adversary can disclose all the secrets of the protocol only by one session of protocol eavesdropping. Finally, we conclude the paper in Sect. 4.

2 Tewari and Gupta authentication protocol

Throughout the paper, we use the notations represented in Table 1, which are similar to the notations used by Tewari and Gupta [20].

The Tewari and Gupta ultralightwight authentication protocol, as depicted in Fig. 1, works as given below:

Fig. 1
figure 1

Tewari and Gupta ultralightweight authentication protocol [20]

Full size image
  1. 1.

    The reader starts the protocol and sends ‘hello’ message to the tag.

  2. 2.

    The tag once received the message, sends its old and new pseudonyms to the reader, i.e., \((IDS_{new}, IDS_{old})\).

  3. 3.

    Upon receipt of the message, the reader searches its database based on received \(IDS_{old}\) and \(IDS_{new}\). If the reader does not find any match, stops the protocol, otherwise it:

    • assuming the tag records in the reader side are \((IDS_{new}, IDS_{old})\) and \((K_{new}, K_{old})\), if \(IDS'_{new}=IDS_{new}\) and \(IDS'_{old}=IDS_{old}\) then \(IDS_{new}\) and \(K_{new }\) are used through calculations;

    • generates two 96-bit random numbers m and n;

    • calculates P, Q and R as below:

      • \(P=IDS\oplus m \oplus n\);

      • \(Q=K\oplus n\);

      • \(R=Rot(Rot(K\oplus n, IDS), K\oplus m)\);

    • and sends (P, Q, R) to the tag.

  4. 4.

    Once reception of the message, the tag:

    • extracts n as \(Q\oplus K\) and m as \(P\oplus IDS\oplus n\);

    • calculates \(R'=Rot(Rot(K\oplus n ,IDS), K\oplus m)\). If \(R'=R\), it authenticates the reader; otherwise, it stops the protocol. If the reader has been authenticated, the tag:

      • calculates \(S=Rot(Rot(IDS\oplus m,K), R'\oplus n)\);

      • sends S to the reader and goes to updating phase.

  5. 5.

    The reader once received the message, calculate \(S'=Rot(Rot(IDS\oplus m, K), R\oplus n)\) with its local values, and if \(S'=S\), then the reader successfully authenticates the tag and goes to updating phase.

  6. 6.

    In the updating phase, the tag and the reader both update their \(IDS_{old}\), \(K_{old}\), \(IDS_{new}\) and \(K_{new}\) as below:

    • \(IDS_{old}=IDS_{new}\)

    • \(K_{old}=K_{new}\)

    • \(IDS_{new}=Rot(Rot(IDS\oplus n, K\oplus n), IDS\oplus m)\);

    • \(K_{new}=Rot(R\oplus n, IDS\oplus m)\);

It should be noted the protocol includes a process to synchronize the tag and the reader records of IDS and K, if the reader has not updated its records of the tag in the last session successfully. However, it has no effect on our attack because we will disclose all secret parameters. Hence, we presented the protocol procedure when the synchronization between the tag and the server remains unbroken.

3 Secret disclosure attack against Tewari and Gupta protocol

Adversary model The attacker in this paper is a passive adversary who is able to only eavesdrop the ongoing reader-tag message exchanged without been detected.

Attack procedure Tewari and Gupta  [20] claim that their protocol is resistant against all known active and passive attacks, including secret disclosure attack. However, we present a rather simple passive attack which can disclose all secrets of the protocol as follows:

  1. 1.

    (Phase 1: Learning Phase:) In this phase of the attack, the adversary eavesdrops one session of the protocol and stores the exchanged messages of the protocol including \(IDS_{old}\), \(IDS_{new}, P, Q, R\) and S.

  2. 2.

    (Phase 2: Passive Secret Disclosure Attack:)

    In this phase of the attack, the adversary by using values which has been eavesdropped in the previous phase, can disclose all secrets of the protocol as bellow:

    1. (a)

      for \(i=0,..., L\), the adversary does:

      • \(S\ggg i \rightarrow x\);

      • \( IDS \oplus x \rightarrow m\);

      • \( P \oplus m\oplus IDS \rightarrow n\);

      • \( Q\oplus n\rightarrow K\);

      • If \(Rot(Rot(K\oplus n ,IDS), K\oplus m)=R\):

        • \(IDS \rightarrow IDS_{old}\)

        • \( K \rightarrow K_{old}\)

        • \(Rot(Rot(IDS\oplus n, K\oplus n), IDS\oplus m)\rightarrow IDS_{new}\);

        • \(Rot(R\oplus n, IDS\oplus m)\rightarrow K_{new}\);

        • returns \(IDS_{old}\), \(IDS_{new}\), \(K_{old}\), \(K_{new}\), n and m.

So, the attacker can disclose all secrets of the protocol only by eavesdropping one session of the protocol and doing the above offline operations which its related code can be executed in a fraction of second in any ordinary personal computer. Given this secret disclosure attack, any other attack such as impersonation attack, desynchronization attack or traceability attack would be trivial.

3.1 Implementation results

We implemented the proposed attack using C++ to verify the correctness of the proposed procedure [6]. For example, for \(L=32\), which means that all parameters are 32-bit variables, consider the following parameters:

figure a

Then the transfered messages that are eavesdropped by the adversary are as follows:

figure b

Now, when we apply our attack, for \(i=26\) we have:

figure c

It is clear that all parameters have been extracted correctly which confirms the correctness of our attack.

4 Conclusion

In this paper, we analyzed the security of an ultralightweight authentication protocol, which had been recently proposed by Tewari and Gupta [20]. We present a passive secret disclosure attack, for which the success probability is ‘1’ and the complexity is only eavesdropping a session of the protocol.

The results of this paper, along with other resent papers on the security of IoT devices such as [18, 19], clarify this fact that we have not met all security desired for protocols related to these devices and it needs more efforts to address them. Hence, we would like to quote the last sentence of the conclusion of [18] as the last sentence of this paper’s conclusion:

We should work together to use the knowledge we gained to protect IoT devices or we might face in the near future large scale attacks that will affect every part of our lives.